Citation
Comparative evaluation of strategic construction-market forecasting methodologies

Material Information

Title:
Comparative evaluation of strategic construction-market forecasting methodologies
Creator:
Fetterhoff, Otto George
Publication Date:
Language:
English
Physical Description:
xiv, 166 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Analytical forecasting ( jstor )
Counties ( jstor )
Economic research ( jstor )
Forecasting standards ( jstor )
Forecasting techniques ( jstor )
Marketing ( jstor )
Mathematical variables ( jstor )
Momentum ( jstor )
Research methods ( jstor )
Statistical forecasts ( jstor )
Building Construction thesis, Ph. D
Dissertations, Academic -- Building Construction -- UF
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 2004.
Bibliography:
Includes bibliographical references.
General Note:
Printout.
General Note:
Vita.
Statement of Responsibility:
by Otto George Fetterhoff III.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
022886282 ( ALEPH )
880637206 ( OCLC )

Downloads

This item has the following downloads:


Full Text











COMPARATIVE EVALUATION OF
STRATEGIC CONSTRUCTION-MARKET FORECASTING METHODOLOGIES
















By

OTTO GEORGE FETTERHOFF III


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


2004
































Copyright 2004

by

Otto George Fetterhoff III

































This dissertation is dedicated to my lovely wife Michele and my two wonderful sons Hans and Alexander. This endeavor would not have been possible without their unwavering patience, support, and love.















ACKNOWLEDGMENTS

First I would like to acknowledge William O'Brien for his support throughout the past four years. As my committee chair, he has provided the needed academic perspective and theoretical balance to what began as a rather pragmatic undertaking. Dr. O'Brien believes that identifying and applying an individual's inherent interests and abilities are a prerequisite to the success of any endeavor.

I would like to acknowledge Marc Smith for serving as cochair of my committee.

His fundamental knowledge of the subject matter and keen perceptiveness were the origin of significant portions of this work. I would like to acknowledge David Ling, Robert Stroh, and Charles Kibert for serving as members of my committee. Their open and insightful feedback was instrumental in assuring a quality and ecumenical outcome.

Finally, I would like to acknowledge Brian Morris and the URS Corporation for making the temporal, financial, and other required resources available to complete this endeavor. Their support helped to minimize the burden of this undertaking on myself and my family.


iv
















TABLE OF CONTENTS

page

A CK N O W LED G M EN TS .............................................................................................. iv

LIST O F TA BLES............................................................................................................. ix

LIST O F FIG U RES .......................................................................................................... xii

A BSTRA CT..................................................................................................................... xiii

CHAPTER

1 IN TRO D U CTIO N ...........................................................................................................1

Issues Leading to this Research................................................................................ 2
Problem Statem ents ......................................................................................................4
Objectives of the Research ....................................................................................... 4
Benefits and Significance of the Research .............................................................. 5
O rganization of this Study ....................................................................................... 6

2 LITERA TU RE REV IEW ............................................................................................ 9

Management Approaches to Strategic Construction Marketing.............................. 9
Strategic M arketing D efinitions ....................................................................... 9
Strategic Marketing Growth Models and Planning Processes ....................... 10
Sum m ary of M anagem ent A pproaches ........................................................... 16
K ey Indicators of Construction A ctivity................................................................. 17
Construction-M arket Segm entation................................................................. 17
Trends and Forces in the M arketplace............................................................ 19
Identification of K ey Indicators....................................................................... 20
Environm ental and Political Indicators ............................................................ 24
K ey Indicator Selection and Constructs ......................................................... 26
Population................................................................................................. 27
G eographic advantage .............................................................................. 27
Initial infrastructure................................................................................. 28
Em ploym ent transition ............................................................................ 29
Econom ic environm ent ........................................................................... 29
Financial resources................................................................................... 29
Sum m ary of the K ey Indicators....................................................................... 30
A pproaches to Construction-M arket Forecasting................................................... 31


v









Qualitative Techniques................................................................................... 31
Quantitative Techniques...................................................................................33
Regression analysis .................................................................................. 35
Dependent regression techniques ............................................................ 36
Interdependent regression techniques .................................................... 36
Trend analysis ......................................................................................... 37
Gap analysis ..............................................................................................37
Law of Universal Gravitation.......................................................................... 39
M omentum analysis .....................................................................................39
M omentum forecasting in the stock market .................................................40
Economic momentum .............................................................................. 41
Summary of Forecasting Approaches ........................................................... 42
Summary of Research Questions............................................................................ 43

3 MOMENTUM THEORY DERIVATION AND APPLICATION..............45

M omentum Theory Introduction ................................................................................45
Differentiation between Momentum and Regression Approaches............47
Data Sample and Data Sources.............................................................................. 48
Momentum Theory Applied to Strategic Construction-Market Forecasting......49
M ethodology for Analyzing County M omentum ............................................ 49
M ass of key indicators...............................................................................52
Velocity of key indicators ....................................................................... 53
Influence of key indicators ....................................................................... 54
M omentum of key indicators .......................................................................55
Total momentum of a county .......................................................................55
Derivation of the M omentum Index ................................................................ 56
M omentum Index Slope ...................................................................................58
Summary of M omentum Theory ............................................................................ 59

4 METHODOLOGY OF COMPARATIVE VALIDATION.......................................63

Alternative Research Approaches.......................................................................... 64
Direct Forecasting ........................................................................................... 64
Factor Analysis.................................................................................................65
M ultivariate-Regression ...................................................................................66
M ERIC Economic M omentum Analysis.............................................................68
Gap Analysis ....... .............................................. .............................................69
Comparative Validation of the Forecasting M ethods............................................ 71
Trend Analysis of Key Construction Indicators .............................................. 71
M ethodology for Comparative Validation .................................................... 72
County Rank Variance........... ........... ......................... ...................................... 73
Cluster Analysis..................................................................................................... 74
Overview ........................................................................................ 74
M ethodology for Cluster Analysis .................................................................. 75
Construction-market classification.......................................................... 76
Key indicators and cluster comparison .................................................. 77


vi









Forecasting m ethods and cluster com parison.......................................... 77

5 K EY IND ICA TOR FIND IN G S................................................................................. 78

Overview of K ey Indicators Results...........................................................................78
Financial Resources.......................................................................................... 80
Econom ic Environm ent ................................................................................... 80
Em ploym ent Transition ................................................................................... 81
Initial Infrastructure.......................................................................................... 81
Population........................................................................................................ 82
G eographic A dvantage ................................................................................... 83
Construction-M arket Classification....................................................................... 84
Key Indicators and County Clusters....................................................................... 88

6 FORECASTING METHODOLOGY FINDINGS .................................................. 91

Statistical O verview .................................................................................................91
One, Tw o, and Three Y ear Trend Projections................................................ 92
V alidation of Statistical Significance .............................................................. 94
A ccuracy of Forecasting M ethodologies ................................................................ 97
D irect Forecasting ........................................................................................... 97
Factor Analysis.................................................................................................98
M ultivariate-Regression ....................................................................................100
M om entum Analysis .........................................................................................100
Gap A nalysis .....................................................................................................102
Missouri Economic Research and Information Center Economic
M om entum A nalysis......................................................................................102
W here are the Best Construction-M arkets?..............................................................103
County Rank V ariance..............................................................................................104
Construction-M arket Classification..........................................................................108

7 CON CLU SIO N S OF TH E STUD Y ............................................................................111

Research Intent and Expectations.............................................................................111
M om entum Theory............................................................................................112
Com parative V alidation of the Forecasting M ethods........................................112
Research Results......................................................................................................113
K ey Construction Indicators..............................................................................113
Forecasting M ethods .........................................................................................115
M ethodology vs. com plexity................................................................. 115
Best construction-m arkets..........................................................................117
Threats to Research V alidity ....................................................................................118
External Validity . --- - ---............. ......... ......................... ..............................118
Construct validity ..................................................................................... 119
Statistical V alidity .............................................................................................120
Internal V alidity.................................................................................................120
Future Research ........ -------------- . . --.-............ ..............................................121


vii









K ey Ind icato rs ...................................................................................................12 1
Environmental and Political Influences.............................................................121
Momentum Forecasting Theory ........................................................................122
Construction-Marketplace Life Cycle ...............................................................122
O ther O pportunities ...........................................................................................124

APPENDIX

A LIST OF POTENTIAL KEY CONSTRUCTION INDICATORS.............................125

B RESEARCH VARIABLE DATA FOR YEARS 1990 THROUGH 2002.................146

LIST OF REFERENCES.................................................................................................160

BIOGRAPHICAL SKETCH ...........................................................................................165


viii
















LIST OF TABLES


Table page

2-1 Summary of construction activity variables listed in Appendix A........................20

2-2 Key indicator constructs and associated variables ................................................27

3-1 Research variable data, descriptions, and sources................................................ 51

3-2 Comparison of momentum theory variable definitibns.........................................52

3-3 Pearson Correlation Coefficient to nonresidential permit activity ........................57

3-4 Total momentum of a county for years 1991 through 2002..................................60

3-5 Momentum index by county for years 1991 to 2002..............................................61

4-1 Six variables used in direct forecasting methodology ............................................65

4-2 Exam ple gap analysis calculation.......................................................................... 70

4-3 Payroll gap type and definitions ............................................................................ 70

4-4 Six forecast methodologies and their twelve associated output variables..............72

4-5 County m arket share classifications ....................................................................... 76

5-1 Key construction indicator constructs and associated variables.............................78

5-2 Pearson correlation results for each research variable............................................79

5-3 Results of the cluster, gap, and market share classification methods.....................85

5-4 Pearson correlation matrix for the three county classification methods.................86

5-5 Results of the key indicator and cluster comparison for the year 2002..................90

6-1 Forecast methodology comparison for 1, 2, and 3 year trend projections..............93

6-2 Forecast methodology F & t statistic comparison for 1, 2, and 3 year trend
projections ...... ..----------------------...................... ................................ 96


ix









6-3 Forecast m ethodology rank and type com parison .................................................. 99

6-4 Rank of nonresidential construction-markets as forecasted by total tax
collections using a 3 year trend ..............................................................................106

6-5 Total county rank variance for forecast years 1996 through 2002...........................107

6-6 Results of forecasting m ethodology and cluster com parison ...................................110

7-1 The relationship between the four construction-marketplace life cycle
stages and m arketing m ix actions...........................................................................123

A-I Econom ic indicators ................................................................................................126

A-2 Construction and infrastructure indicators...............................................................128

A-3 Safety and health indicators.....................................................................................131

A-4 Education, social, and governm ent indicators .........................................................133

A-5 Air environs ental indicators ...................................................................................136

A-6 W ater environm ental indicators...............................................................................138

A-7 Land environm ental indicators ................................................................................140

A-8 Ecology environm ental indicators ...........................................................................141

A-9 Sound environm ental indicators ..............................................................................142

A-10 Natural resource environm ental indicators............................................................143

B-1 1990 variable data....................................................................................................147

B-2 1991 variable data....................................................................................................148

B-3 1992 variable data....................................................................................................149

B-4 1993 variable data....................................................................................................150

B-5 1994 variable data................................................................................................. 151

B-6 1995 variable data....................................................................................................152

B-7 1996 variable data.................. ................................. ............................................153

B-8 1997 variable data....................... ........................... ............................................154

B-9 1998 variable data.--------------.. . ----........... ........................ .................................155


x









B-10 1999 variable data..................................................................................................156

B-11 2000 variable data..................................................................................................157

B-12 2001 variable data..................................................................................................158

B-13 2002 variable data ..................................................................................................159


xi
















LIST OF FIGURES


Figure page

1-I O rganization of this study........................................................................................ 6

3-1 Map of the 67 counties in the State of Florida .......................................................50

3-2 Seven general steps of momentum analysis ...........................................................52

3-3 Three-dimensional relationships of momentum.....................................................56

3-4 Momentum index of each Florida County plotted for years 1991 to 2002 ............62

4-1 Four steps for comparative validation of research methods .................................. 63

5-1 Maps of county classification analysis .................................................................. 86

6-1 Map of the 2005 total tax-direct forecast results using a three year
trend projection ......................................................................................................105

6-2 County rank variance comparison for the year 2002................................................107


xii















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy COMPARATIVE EVALUATION OF STRATEGIC CONSTRUCTION-MARKET FORECASTING METHODOLOGIES By

Otto George Fetterhoff III

May 2004

Chair: William J. O'Brien
Cochair: Marc T. Smith
Major Department: Building Construction

The relative historical stability of the U.S. economy and its strong influence on the construction industry have allowed large U.S. design and construction firms to naturally grow and adapt to the slow changes of the construction-market. But business changed suddenly for many large firms in 2001. The economic recession, combined with the market fallout from the events of September 11, 2001, delayed or cancelled many design and construction projects. Despite these historic events, investor expectations prevailed, sending these firms searching to find new and alternative markets.

This research sought to examine many of the challenges a marketing professional is confronted with when searching for new competitive markets. These challenges generally include questions regarding the key predictive indicators of construction activity, the methods of market forecasting, and the classification and selection of new potential markets. This research was also intended to initiate decisions regarding the spatial structure of a large design and construction firm.


xiii








A review of the literature provided many different management approaches to strategic construction-market forecasting. Over 250 different indicators (in over 56 categories) were identified that could potentially influence construction activity within a given market. The literature review also provides an overview of the variety of forecasting methodologies that can be used to forecast and classify construction activity.

Historical data were collected for all of Florida's 67 counties for the period 1990 through 2002. A new momentum forecasting methodology is presented that was derived from Sir Isaac Newton's three laws of motion. This unique forecasting approach (along with five other existing approaches) is used to forecast construction-market activity at the county level.

The best key indicators of future nonresidential construction activity at the county level were found to be the total annual tax collections, the total annual assessed value of commercial land, and the total annual number of wage and salary jobs in a county. The current industry practice of using housing starts as a key indicator was found to be one of the least accurate indicators of nonresidential construction activity in the State of Florida relative to the other methods tested.

A direct forecasting methodology (using a single variable) was found to be the most accurate predictor of future nonresidential construction activity. This was the simplest of all the forecasting methodologies used, and consistently outperformed the more complex statistical multivariate-regression techniques. The new momentum methodology was found to be a highly accurate alternative forecasting methodology that is less complex but more meaningful than the traditional statistical regression approaches.


xiv













CHAPTER 1
INTRODUCTION

The general domain of the following research is market forecasting in the U.S. construction industry. More specifically, our research is intended for large design and construction firms working in the nonresidential construction-market. Our research applied several existing and one new forecasting methodology to the difficult management task of analyzing and prioritizing multiple construction-markets. Our research is intended to help these large firms focus their resources on the markets with the most opportunity. Geographical coverage is generally seen as a strategic business decision. Marketing and sales are organized mainly as a business process. Our research is intended to initiate decisions regarding the spatial structure of the organization. In short, the research goal is to locate the opportunity, not win an opportunity.

A new forecasting method is presented that integrates Newton's natural science theory of momentum. This unique forecasting approach (along with several other existing approaches) is used to estimate nonresidential construction activity in all 67 counties of the State of Florida. A 13 year time period was used for our study that began 1990 and ended in 2002. This momentum forecasting methodology is introduced as a more understandable way to conceptualize the complexities of strategic market forecasting. Finally, our research completed a comparative validation of the various forecasting techniques to confirm the accuracy of the different approaches.


1






2


Issues Leading to this Research

A cowboy heading westward came across an Indian lying in the middle of the
wagon trail with his ear to the ground. The cowboy stopped and asked the Indian
what he was listening to. Without getting up and his ear still to the ground, the Indian replied, "covered wagon with two calico horses heading west, large man
with hat and rifle, woman in blue dress and two screaming children, wagon fully loaded with cast iron stove tied to back of wagon." Amazed at the details of the
Indian's prediction, the cowboy asked the Indian how he could describe the wagon with such accuracy. The Indian replied, "Wagon ran over me thirty minutes ago."
- Author Unknown'

Due to the relative historical stability of the U.S. economy and its strong influence on the construction industry, large design and construction (D&C) firms have naturally evolved, or organically grown and adapted with the slow changes of the constructionmarket. But business changed suddenly for many large U.S. D&C firms in 2001. The economic recession combined with the market fallout from the events of September 11, 2001, further delayed or in some cases cancelled the start of many construction projects in markets that were once thought to be secure. Despite the magnitude of these historic events, large D&C firms were soon pressured by their investors to return to the growth they knew before 2001.

These investor expectations forced many large D&C firms to evaluate alternative market areas that were historically not considered by their offices. The market area covered by a local D&C office may include a city, counties, a state, or multiple states. A market can be defined as a geographic or political boundary where D&C firms compete to provide the same services.




1Pastor Jim Henry, First Baptist Church of Orlando (personal communication, August 31, 2003).
2 Hillebrandt, P. M. (1974). Economic Theory and The Construction Industry. London, Great Britain: The Macmillan Press, Ltd. (p. 37).






3


Market Research Services of Florida conducted a study of 37 construction firms with $2 million to $215 million in construction volume and found that most CEO's do not truly believe they are in control of their growth.3 By reviewing growth histories, it was found that 70% were basically reactive marketers or dependent on the marketplace to dictate where, when, how, and if the company would grow. In specialty areas or markets of rapid growth, many reactive firms were successful until that point when the market or geographic area leveled off or declined.4

Construction organizations need a continuity of activity, not only to keep their resources fully employed but to generate a return on investment that will attract the necessary capital for their continued existence and growth.5 Corporations seem to grow (increase in volume) or die. Few are able to remain at a constant level, no matter how hard they try. There is an inevitable need to expand, improve, or increase [a firm's] share of the market.6

Before a large D&C firm begins the process of getting in the door of a client, or

even deciding on whose door to enter, a firm needs to know where the doors are. For the most part, operating companies are expected to find their own growth areas.7 The question then becomes where are the best future markets for a large D&C firm to



3 Pickar, R. L., AGC Construction Marketing Committee. (1995). A Contractor's Guide to Focus Sales and Increase Profitability: A Marketing Workbook for Contractors. Washington, D.C.: Associated General Contractors of America. (p. 69).

4 Pickar, R. L. (pp. 69-70).

s Gerwick, B. C., & Woolery, J. C. (1983). Construction and Engineering Marketingfor Major Project Services. New York: Wiley. (p. 2).
6 Gerwick, B. C., & Woolery, J. C. (p. 19).

Hillebrandt, P. M., & Cannon, J. (1990). The Modern Construction Firm. London: Macmillan. (p. 64).






4


compete? The conventional wisdom is that D&C firms must go where the money is. The money is logically thought to be in the largest population centers and/or the markets closest to these population centers. In other words, start with a major city and target the perimeter counties. While there may be much truth and common sense to this approach, it is simplistic and not founded in proven theory of how a county's construction activity actually develops. While population and adjacency to major cities may be good rules of thumb, our research proposes that there may be better methodologies for strategically forecasting future market opportunities.

Problem Statements

Many questions must be answered by the marketing professional when tasked with finding new competitive markets for a D&C firm. Some of these questions may include;

* Where are the best new markets for a large D&C firm?

* What are the key indicators that best predict a potential market's construction
activity?

* What methodologies are available to identify these new markets? " Are complicated forecasting methods really more accurate than a simple, more
direct, approach?

* Which forecasting methods are the most accurate? " Finally, how should the various levels of market opportunity be segregated so the
highest potential markets can be prioritized and pursued?

Objectives of the Research

A good hockey player plays where the puck is, a great hockey player plays where the puck is going to be. - Retired hockey legend Wayne Gretzky




8 BrainyMedia.com. Wayne Gretzky Quotes. Retrieved February 17, 2004, from http://www.brainyquote.con-/quotes/authors/w/waynegretzky.htnil.





5


The overall objective of our research is to generate knowledge that will assist larger D&C firms with selecting the best future construction-market opportunities based on key construction indicators. While the objectives of our research are to find answers to the strategic marketing research questions listed above, another important objective of our research is to develop and apply the logic of Newton's momentum theory to constructionmarket forecasting. The proposed momentum forecasting methodology will be used to convert key construction indicators and their associated levels of influence into a usable market-opportunity index for all of the counties within the State of Florida.

Benefits and Significance of the Research From an academic perspective, the most significant contribution of our research is the identification and analysis of the key indicators of nonresidential construction activity. The analysis of these key indicators will provide a better understanding of which (and what type of) key indicators best predict a county's construction activity.

The second most significant contribution of our research is applying the concept of momentum theory to construction-market forecasting. The proposed momentum forecasting approach was designed to be less complex but more meaningful than traditional statistical regression approaches. One of the differences of the momentum approach is that each variable includes three-dimensions. These dimensions include the relative size, rate of change, and influence of the variable. These three dimensions are combined into one new variable value (momentum). The momentum forecasting theory is intended to be a more understandable framework that can be used to think about, and conceptualize, the complexities of construction-market forecasting.






6


Another contribution is the application and comparison of various methods of

construction-market classification. Finally, the large firm, nonresidential market focus is another unique contribution of our research.

From an industry perspective, the knowledge gained from our research is useful to large D&C firms for several reasons. Their marketing resources could be more efficiently used to (a) identify developing competitive geographic markets such as individual counties, or county clusters or networks; (b) compare and prioritize these markets; and (c) better predict construction activity in these markets several years in advance.

Organization of this Study

Our research is divided into seven chapters (Figure 1-1). Chapter 1 includes a discussion of the issues leading to our research, statements of the problems to be researched, objectives of the research, and the significance of the research. The following paragraphs describe the content and organization of Chapters 2 through 7.

Chapter Chapter

3 5
Chapter Chapter Momentum Key Chapter
Methodology Iniatr
Introduction ~ Literature -Chapter - Z/ Chapter -ZReview 46Conclusions
Comparative Forecasting
Validation Methods
Methodologies Findings
Figure 1-1. Organization of this study.


Chapter 2 includes an overview of the management approaches to strategic

construction marketing and is followed by a detailed discussion of the key indicators known to forecast the potential for construction activity. Finally, the various potential






7


approaches to construction-market forecasting are reviewed. Several of these different forecasting approaches are selected for use in our research.

Chapter 3 develops and applies the proposed momentum theory to strategic

construction-market forecasting. Next, a discussion of what differentiates momentum analysis from other forecasting techniques is included. The research data sample and data sources are reviewed. Finally, the procedure for analyzing county momentum is detailed and a momentum index is derived.

Chapter 4 discuses all of the steps taken during our research to comparatively

validate the proposed momentum forecasting methodology to five alternative forecasting approaches selected from the literature review. A one, two and three year trend projection analysis is completed using all of the forecasting methods. The forecasted results from all of the techniques are comparatively validated and rank ordered against a county's actual construction activity. Finally, cluster regression analysis is used as a way to group the counties and is compared to two other classification techniques.

Chapter 5 begins with a review of the overall results for the key indicator analysis, and is followed by a discussion of the specific findings for each key indicator variable. Next, results from the cluster analysis and the two other construction-market classification methodologies are reviewed. Finally, findings regarding the relationships between the key construction indicators and county clusters are presented.

Chapter 6 begins with a review of the overall results for the forecast methods including a discussion of the relevant statistics. This is followed by a review of the finding for each individual forecast methodology. The best forecasted future nonresidential construction-markets in the State of Florida are identified. Next, variance






8


results between the projected county rankings and actual county rankings are presented. Finally, findings regarding the relationships between the forecasting methodologies and county clusters are reviewed.

Chapter 7 presents what was found and learned by our research. This chapter begins with a review of the research intent and expectations and is followed by a summary of the research results. The overall threats to the research validity are discussed. Finally, several opportunities are identified regarding future extensions and use of the presented research.














CHAPTER 2
LITERATURE REVIEW

This chapter provides a summary of the literature review completed for our research and is organized into three primary sections. The chapter begins with an overview of the management approaches to strategic construction marketing and planning. Next, key indicators of construction activity are identified, grouped and reviewed. Finally, potential methodological approaches to construction-market forecasting are reviewed and several methodologies are selected for use in our research.

Management Approaches to Strategic Construction Marketing Strategic Marketing Definitions

The words strategic and marketing have been given a variety of definitions

throughout the literature. The following definitions are provided to give a common point of reference for this terminology.

Strategic: Necessary to or important in the initiation, conduct, or completion of a
strategic plan. Of great importance within an integrated whole or to a planned
effect.'

Marketing: The process of planning and executing the conception, pricing,
promotion, and distribution of ideas, goods, and services to create exchanges that
satisfy individual and organizational objectives.2






Merriam-Webster Online Dictionary. Retrieved February 2, 2004, from http://www.m-w.com/cgibin/dictionary.
2 Bennett, P. D. Dictionary ofMarketing Terms (2nd ed.). Lincolnwood, IL: NTC Publishing Group, 1995. (p. 166).


9






10


Strategic Marketing Growth Models and Planning Processes

The following sections of this chapter provide a summary of several business growth models and marketing planning processes that are specific to the design and construction industry and to our research. The intent of these sections is to show where our research fits into the marketing planning process.

One of the most conventional views of marketing and strategic planning for the

design and construction industry comes from BNI Building News (2000). BNI begins by differentiating between sales and marketing.

Sales is closing the deal, the specific project or program. It's the signing of a
contract, the exchange of money for services. Everything up to that point is
marketing.3

BNI differentiates between a marketing strategic plan and a marketing business plan.

A strategic plan is a three to five year road map, whereas the marketing business
plan is a one year increment of that plan.4

BNI then presents a five-step strategic planning process.5 These five general steps include;

* Research and analysis of internal strengths and weaknesses and external
opportunities and threats
* Collective decision making
" Organizational engineering and communication
* Implementation of the plan
" Evaluation and results

Our research would be used during step one of BNI's strategic planning process. More specifically, our research focuses on the external opportunities of the large design 3 BNI Building News Society for Marketing Professional Services. (2000). Marketing Handbookfor the Design & Construction Professional. Los Angeles: BNI Building News. (p. 13).
4 BNI Building News. (2000). (p.27).

5 BNI Building News. (2000). (p. 28).






11


& construction (D&C) firm. BNI describes the research and analysis of these external opportunities as

The identification of external factors, determined by analyzing broad based social,
demographic, cultural and economic trends for marketing and business implications, as well as specific possibilities presented by client/market
requirements.6

Another conceptual framework for strategic marketing and planning comes from Smyth (2000). Smyth presents four possible market-analysis viewpoints for the construction-marketplace. These market views are based on interrelationships among the origin of the view (internal or external) and the direction of the view (top down or bottom up). Our research emphasizes the top down market research view that is external to the organization.

Smyth also presents a nine-step strategic planning process. These nine general steps have similar attributes to those presented previously in the BNI model and include

* Business objectives
* Marketing audit (the Four Views)
* SWOT analysis
* Assumptions
* Marketing objectives and strategies " Estimate expected results
* Identify alternative approaches
* Implementation plan
* Monitoring plan







6 BNI Building News. (2000). (p. 29).

7 (Smyth, H., & NetLibrary Inc. (1999). Marketing and Selling Construction Services. Malden, MA: Blackwell Science. (p. 50)).

8 Smyth, H., & NetLibrary Inc. (1999). (p. 52).






12


Our research would be used in the business objectives step of Smyth's planning process. More specifically, our study focuses on the business objectives relating to the geographic structure of the D&C firm.

Smyth then outlines three traditional models for design and construction firm growth. These growth models are listed below. Our research would be used with Growth Model 2.

" Model 1. Expansion into existing markets
* Model 2. Expansion into new markets " Model 3. Expansion by takeover or merger

Another conceptual framework for strategic marketing and planning comes from the Associated General Contractors of America (AGC 1995). Similar to BNI Building News, the AGC differentiates between a construction marketing approach and sales approach. The AGC's six step marketing approach9 is outlined below. Our research would be applied during the AGC's research and screening task.

0 Research and screen
* Select target
" Advertising, public relations, and networking
* Trust
* Sales
* Close

The AGC then outlines a ten-step marketing process'0 similar to those discussed above. This process is designed to allow the construction firm to target particular markets and decision makers. The steps of this process are listed below.


9 Associated General Contractors of America., & AGC Construction Marketing Committee. (1995). A Contractor's Guide to Focus Sales and Increase Profitability for the Associated General Contractors of America: A Marketing Workbookfor Contractors. Washington, D.C.: Associated General Contractors of America. (p. 9).

" AGC. (1995). (p. 17).






13


" Determine your mission
" Set goals
" Perform internal analysis " Perform external analysis " Establish marketing goals " Generate strategies
* Research and refine strategies " Create and refine promotional and sales tactics " Implement the plan
" Evaluation of results

The AGC also provides a good explanation of an external market analysis:

An external analysis examines the trends in the marketplace: hot-vs.-cold markets,
local economic outlook, market types, available financing, and market needs.
During the external analysis (in the up-and-down, cyclical construction arena) it is important to research basic factors that can create or eliminate a market place for a general contractor. These factors include the competition; the economic, social and
political changes in the marketplace; and the need for particular infrastructure,
facilities, and contracting services."

The AGC then provides a list of the most common reasons that geographical

expansion is initiated by a D&C firm. These reasons are listed below.

" A leveling off of the need in an existing market.
* An increase in competition beyond the rate of area or market growth.
* A major increase in price sensitivity.
* There is a potential long-term downturn in the existing market. " The local area is historically very cyclical. " The local area is dependent on too limited a market for economic health.
* There are strong management resources looking for an opportunity of personal
growth in a more independent atmosphere.

Another conceptual framework for strategic marketing and planning comes from

Gerwick & Woolery (1983). First, Gerwick & Woolery recommend several ways for a


12 AGC. (1995). (p.60).


" AGC. (1995). (p. 20).






14


construction firm to increase its contract volume.'3 These recommendations are similar to Smyth's growth models discussed above and are listed below. " Geographical expansion into new market areas.

* Greater market penetration or increasing the percent share of a firm's existing
market.

* Diversification of a firm's services on its own or through the acquisition of other
firms.

* Increase a firm's scope of services within a firm's area of expertise.

The focus of our study is on the first recommendation, geographical expansion into new markets.

Gerwick & Woolery outlined a six-step marketing plan.'4 The six primary steps of this approach are listed below.

* Establishing the firms long range objectives " Evaluating alternative marketing plans
* Selecting a tentative plan for use
* Implementing the plan
" Monitoring the firm's performance while using this plan
* Revising the marketing plan as necessary to achieve the firm's desired objectives Our research would be used during step one of this six-step marketing plan.

Still another conceptual framework for strategic marketing and planning comes

from Friedman (1984). Friedman begins by differentiating between construction-market forecasting and market planning.

Planning is determining what a company wants in the future and developing
methods to achieve it. Forecasting describes the type of external environment that
can be expected.15


1 Gerwick, B. C., & Woolery, J. C. (1983). Construction and Engineering Marketingfor Major Project Services. New York: Wiley. (pp. 38-39).
1 Gerwick, B. C., & Woolery, J. C. (1983). (p. 38).






15


Friedman presents a six-step sales process16 for the construction industry similar to those previously presented. These steps are listed below. " Establishing corporate marketing objectives
" Generating project leads
" Qualifying prospects
" Conducting sales interviews
" Preparing proposals & presentations
* Entering into contract negotiations and closing

Our research would be used during step one of Friedman's process.

Finally, a more complex systems approach to strategic construction marketing was presented by Fisher (1986).17 Fisher's concept was based in part on the work of Adler (1967). 18 This systems approach embraces the marketing complexities of the organization and its interface with its environment through a series of input-output diagrams, interaction maps, data flow diagrams, and an overall marketing information system diagram.'9 While this approach offers an alternative way of looking at market planning, the models are; complex, time consuming to construct, and difficult to adapt to different situations; the models may also provide inaccurate results due to the lack of hard data; and finally they are perceived as too restrictive by marketing executives.0





15 Friedman, W. (1984). Construction Marketing and Strategic Planning. New York: McGraw-Hill. (p. 12). 16 Friedman, W. (1984). (p. 142).

1 Fisher, N. (1986). Marketing for the Construction Industry: A Practical Handbook for Consultants, Contractors, and other Professionals. London: Longman; J. Wiley. 18 Adler, L. (1967). Systems Approach to Marketing, Harvard Business Review. Sept./Oct. (pp. 105-118).

9 Fisher, N. (1986). (p. 55-67, 115).
20 Fisher, N. (1986). (p. 66).






16


Fisher also presented a ten-step marketing planning approach.2' These ten steps are listed below. Our research would be used during steps 2 through 4. " Internal company appraisal.
" External company appraisal. " List existing, and identify new, business opportunities. " Assess future market potential for each opportunity identified. " Assess company ability to secure successful business in identified market sectors. " Rank options based on potential profit yield, prevailing conditions and key skills
needed (existing or new).
" Define and agree on marketing objectives. " Prepare detailed business plan for meeting objectives. " Putting the market plan into action. " Monitor and review plans in light of conditions encountered and performance
achieved.

Summary of Management Approaches

There are probably as many ways to develop a strategic marketing plan as there are design & construction firms in the United States. What was learned from the preceding literature review is that a D&C firm must systematically plan for what the firm wants to be in the future, and chart a course to achieve that vision.

Strategic planning often yields new undertakings for the firm in the form of new geographic locations and offices, mergers and acquisitions, and start-ups of new
businesses.

While BNI, Smyth, AGC, Gerwick & Woolery, Friedman, and Fisher all offer a well structured set of tasks, conceptual models, and definitions in regards to strategic construction-market planning, they stop short of offering a detailed methodology on how to actually implement the research, prioritize, and select a potential new constructionmarket. Another observation is that the available construction marketing literature is



2' Fisher, N. (1986). (p. 133).

2 BNI Building News. (2000). (p. 33).






17


focused on the sales aspects of constructing marketing for smaller firms, not the identification of new competitive markets for larger design and construction firms.

This research is intended to be used as a top down appraisal tool for identifying

new external business growth opportunities, and assessing the future market potential for each opportunity identified. Our research is intended to be applied in the initial stages of a firms marketing planning process, during the tasks of research and screening.

Our research is intended to initiate decisions regarding the geographical expansion and spatial structure of the organization. In short, the research goal is to locate the opportunities, not win an opportunity.

Key Indicators of Construction Activity The second area of literature review for our research is the identification of key indicators of construction activity. The following sections of this chapter begin with a brief discussion of construction market segmentation. Next, a large number of potential key construction indicators are identified and reviewed. Finally, several of these indicators are selected for use in our research and are classified into groups to allow a more understandable discussion of their relationship to construction activity. Construction-Market Segmentation

Berkowitz et al. (2000) defines market segmentation as the process of "aggregating prospective buyers into groups that (a) have common needs and (b) will respond similarly to a marketing action. The groups that result from this process are market segments, a relatively homogenous collection of prospective buyers."23 BNI Building News (2000)24 provided a similar definition for market segmentation. 23 Berkowitz, E. N. et al. (2000). Marketing (6th ed.). Boston Massachusetts: Irwin McGraw-Hill. (2000). (p. 13).






18


Thomsen (1989) classifies construction-markets in three primary segments. These segments include; (a) the construction-type markets (e.g., transportation, power, buildings), (b) the geographic type markets (e.g., U.S. market, Florida market, Orlando market), and (c) the service markets (e.g., architecture, engineering, construction, construction management).2 Gerwick and Woolery (1983)26 outline a similar market segmentation approach. Finkel (1997) provides a higher macro classification of the type of construction-markets. Finkel's market segments include; (a) private residential, (b) private commercial (nonresidential), and (c) public construction. Engineering News Record (ENR) uses similar methods to segment construction-markets. ENR's eleven type of work classifications27 are listed below.

* Building
* Manufacturing
* Industrial
" Petroleum
" Water
* Sewer & Waste " Transportation
" Hazardous Waste
* Power
" Telecommunications " Other

Smyth (1998) recognized the process of market segmentation is not as clear-cut as it appears. Public and private owners build both residential and nonresidential projects in



2 BNI Building News. (2000). (p. 17). 25 Thomsen, C. (1989). Managing Brainpower: Organizing, Measuring Performance, and Selling in Architecture, Engineering, and Construction Management Companies. Washington, D.C.: American Institute of Architects Press. (p. 13). 2' Gerwick, B. C., & Woolery, J. C. (1983). (p. 20).

2 Tulacz, G., & Powers, M. (May 19, 2003). The Top 400 Contractors. ENR. (p. 63).






19


almost any geographic location. To better capture these relationships, Smyth developed a simple market segmentation model.28 This model graphically shows the relationships between the traditional market segments in the context of the overall marketplace.

Due to the cost and availability of research information, and to our research scope and schedule constraints, a broader more general market segmentation approach is taken for our research. Our research focuses on the public and private construction activity in the nonresidential construction-market at the Florida county level. Trends and Forces in the Marketplace

Regardless of the market classification method chosen by a D&C firm, all of the market segments within the market place, and the market place itself, are shaped by changes in the marketing environment. These changes in the marketing environment are caused by uncontrollable trends and forces that are external to the D&C organization. Berkowitz et al. (2000) indicate that these environmentalforces involve social, economic, technological, competitive, and regulatory changes.29 These changes in the marketing environment are a source of opportunities and threats to be managed.

The marketing literature uses the term environmentalforces to describe the various changes within the marketplace. In the design and construction industry, the term environmental is more closely associated with the natural and physical sciences (i.e., natural environment and environmental sciences). To avoid any confusion, our research




Smyth, H. J. (1998). Innovative Ways of Segmenting the Market: Practice Guide No. 1. Oxford: Oxford Brookes University, Center for Construction Marketing. (p. 12). 29 Berkowitz, E. N. et al. (2000). (p. 13).

30 Berkowitz, E. N. et al. (2000). (p. 74).






20


uses the term key indicators, a term more conventional to the construction industry, to describe the forces and trends that are changing in the marketplace. Identification of Key Indicators

A review of the construction economics, construction marketing and

macroeconomic literature has provided a wide variety of sources that identify the indicators that could potentially influence construction activity within a given market. The indicators found in this literature review are summarized in Appendix-A. Appendix A includes over 250 indicators in 56 different categories, the potential units of measurement, an example scale (or index), and potential data sources. Table 2-1 summarizes the quantity of indicators and indicator categories found in Appendix A. It should be noted that many of the variables (e.g., Environmental and political variables) outlined in Appendix A are not used in our research for a variety of reasons. These reasons are outlined later in this chapter. The following paragraphs summarize several of the typical sources that were used to generate these key indicators.


Table 2-1. Summary of construction activity variables listed in Appendix A.
Variable categories Quantity of variables
Economic, construction, and infrastructure 6 84
Community, government, and politics 8 82
Environmental 42 86
Total: 56 252


A comprehensive list of international construction indicator sources was published by the Organization for Economic Co-operation and Development (OECD).31 The publication includes a detailed list of the public and private sources of data for the main

3 Organisation for Economic Co-operation and Development. (2002). Main Economic Indicators: Comparative Methodological Analysis : Industry, Retail and Construction Indicators. Paris: Organisation for Economic Co-operation and Development. (p. 65-78).






21


economic indicators of construction in 30 different countries. The OECD also lists the sources for the indicators offuture and actual construction activity in the same 30 member nations. For its source of U.S. construction data, the OECD used the FW Dodge Corporation forfuture construction activity indicators, and the U.S. Census Bureau as the source for actual construction activity indicators.

In 1999, Standard & Poor's DRI (F.W. Dodge Division of McGraw-Hill

Construction Information Group) published Building New Markets: Global Construction Market Opportunities and Risks.2 The purpose of this prospectus was to sell international construction-market research information services to large D&C firms. In Standard & Poor's Method and Analysis summary, the market forces affecting the construction-market in the short-term and the implications of long-run structural changes on construction-markets are discussed.33 The following paragraph is a brief summary of the construction indicators discussed in the S&P DRI prospectus.

In the short-term, the growth of GDP, interest rates, inflation, unemployment, international trade and financial linkages and exchange rates can affect constructionmarket demand. Other key indicators listed include domestic investment on commercial structures, dwellings, infrastructure and the country's ability to finance its projects. Long-run structural changes that influence a country's construction-market were subgrouped by Standard & Poor's into three broad dimensions. These dimensions include

(a) shifts in the sectoral economic activity, (b) changes in the pattern of urbanization; and



3 Standard & Poor's, DRI & F.W. Dodge. (1999, December). Building New markets: Global Construction Market Opportunities and Risks (Prospectus for Multi-Client Study). Lexington, MA: The McGraw-Hill Companies Construction Information Group.
3 Standard & Poor's, DRI & F.W. Dodge. (1999, December). (pp.' 1-13).






22


(c) the demographic transition. Systematic sectoral shifts in economic activity were said to be the most influential and include movement from low to high incomes, and movement from agriculture through manufacturing into a service sector economy. The next most influential change on construction-markets is urbanization and includes a country's transition from widely dispersed small communities to large urbanized cities. As transportation services expand, secondary urban areas develop around the large cities. Demographic changes are the last most influential changes that affect a country's construction activity and include; population size, population growth, fertility and mortality rates, household size, education levels, and per capita income.

The North American Construction Forecast (NACF) published a report in 2002 on the national U.S. construction activity outlook for the upcoming 2003 year.34 In this report Ken Simonson, Chief Economist of the Associated General Contractors, was interviewed and said that there are two key indicators of U.S. construction health: construction employment and value put in place. Simonson went on to identify several key indicators in specific segments of the construction-market. These indicators are listed below.

" Interest rates and unemployment for single family home construction. " Tax receipts, bond referendums and property values for the public construction
sector.

" Consumer spending on homes, automobiles and health related items for the broad
private nonresidential segment, (e.g., building supply store construction, auto sales
facilities, drug stores and health care facilities).



3 Wright, R. (2002, October 16). U.S. Construction Activity Stagnant, but Promises Gradual Improvement; Government-Related Construction Shows Diminished Activity in 2003. North American Construction Forecast. Retrieved October 1, 2003, from http://www.nacf.com/simonson 02.html.






23


* Unemployment, capacity utilization and profits for factory, office, warehouse,
business related hotel, restaurant and car rental agency construction.

Steele Analytics published Construction and Real Estate Market Pulse (2003)35 on

their website that lists eleven indicators for the U.S. construction and real estate market.

These indicators are listed below.

* Building permits
* Construction employment
* Construction equipment producer price index
* Construction equipment shipments " Construction spending " Median sale price of existing single family homes " Existing single family homes sales " Lumber producer price index
* Median sale price of new single family homes " New single family homes sales
* Housing starts

The New Jersey Construction Reporter36 (a publication of the Division of Codes

and Standards, New Jersey Department of Community Affairs) uses four major

construction indicators to evaluate the activity of the New Jersey construction industry.

These construction indicators include a quarterly comparison of;

* Estimated cost of construction " Authorized housing units
* Authorized office space " Authorized retail space

The Metropolitan Washington Council of Governments published the report

Economic Trends and Commercial Construction Indicators for Metropolitan Washington



35 Construction and Real Estate Market Pulse. Steele Analytics. Retrieved October 1, 2003, from http://www.steeleanalytics.com/construction.htm. 36 New Jersey Construction Reporter, March 2003 Highlights. Division of Codes and Standards, New Jersey Department of Community Affairs. Retrieved October 1, 2003, from http://www.state.nj.us/dca/codes/cr/subform.shtm.





24


(2003).37 This report identified twelve regional construction indicators that were used to

forecast construction activity in the Washington D.C. area. These twelve regional

indicators are listed below.

* Population
* Employment
* Federal spending
* Labor force
* Construction
" Mortgage rates
* Home sales
* Housing related inflation
* Inflation
" Income
" Retail sales
* Airline passengers

Other sources for U.S. construction-market forecasting indicators and

methodologies in use today are listed below.

" Construction Review (U.S. Department of Commerce)
* County Business Patterns (Bureau of the Census) " Survey of Current Business (U.S. Department of Commerce)
* Bureau of Labor Statistics (U.S. Department of Labor) " Construction Online McGraw-Hill
* Reed Construction Data
* Associated General Contractors
* Lend Lease Real Estate Investments " PricewaterhouseCoopers

Environmental and Political Indicators

There are as many reasons for including or excluding political and natural

environmental indicators as there are variables in Appendix A. I have attempted to

summarize three primary reasons why these variables have been avoided in our research.


37 Economic Trends and Commercial Construction Indicators for Metropolitan Washington. Metropolitan Washington Council of Governments. Retrieved October 1, 2003, from http://www.mwcog.org/uploads/comm5ittee-documents/9FtYXw2O3O715l44112.ppt.






25


These reasons include; (a) the problem of defining what the natural or political environment actually is, (b) the requirements for assessing the political and natural environment, and (c) the intended purpose of our research.

First, the National Environmental Policy Act of 1970 (NEPA) mandates that

federal government agencies assess the environmental impacts of actions "which may have an impact on man's environment."38 Most state and local government agencies have adopted NEPA's regulatory framework for use at their levels. But the meaning of a man 's environment was not defined by this legislation nor has it been over 30 years later. Is the natural environment defined by wooded scenes, fresh air, clean water, low noise levels, and a pleasant suburban neighborhood? Or is it the dynamics of the food chain, endangered species, agricultural pesticides, and global warming? Is the political environment defined by the availability of health care, quantity of homeless persons, free child day care, and the percent of voters registered Democratic? Or is the political environment defined by public safety, the quality of public transportation, classroom student to teacher ratios and economic stability? These are all interrelated characteristics of a man's environment that may be impacted differently depending on the specific action taken. Further, all the variables listed in Appendix A can be made political. While it is generally agreed that the elements of the political and natural environment include the aesthetic, historic, cultural, economic and social aspects of a community, "the ultimate selection of what is really important in any one case is very much an art."39




38 National Environmental Policy Act, Title I, Sec 102(2) (A). 39 Jain, R.K. et al. (2002). Environmental Assessment. New York: McGraw-Hill. (p. 5).






26


Second, since our research is being conducted at a strategic level (i.e., county

level), not at a project level, it is difficult to perform an actual assessment of the political and natural environment at the county level for several reasons. First, it is necessary to have a complete understanding and definition of the proposed construction action or project. Next, it is necessary to have a complete understanding of the surrounding environment being affected by the action caused by the location specific nature of political and environmental variables (e.g., air quality, water quality, noise levels, threatened species, public safety, health, social environment). Finally, the defined action or project would have to be combined with the setting to determine the interaction and changes that may occur. For an anti-growth or NIMBY organization to build up any type of resistance, they must have something to resist against. For these reasons, an assessment of the political and natural environment is normally conducted only at the project specific level.

Key Indicator Selection and Constructs

After the initial variables in Appendix A were identified, the preceding literature review was used to narrow the quantity of variables down to approximately 20 to 30. At this point, data availability became one of the primary determinants of which variables were finally selected. A total of sixteen key construction indicators were selected from the previous literature review and were used as independent variables in our research. Each of these sixteen variables including the variable name, type, unit, description, and source are outlined in Chapter 3.

To simplify the discussion regarding the relationships between the key construction indicators, the indicators were grouped into six independent variable constructs. The independent variables and their associated constructs are shown in Table 2-2.






27


Table 2-2. Key indicator constructs and associated variables.
Population Geographic Initial Infrastructure Employment Economic Financial
Advantage Transition Environment Resources
Total Population Proximity- Daily Vehicle Miles Construction Total Total Taxes Population Density Size Factor Traveled Payroll Employment Assessed
Centerline Miles of Total Personal Gross Sales Commercial Roadway Income Price Level Land Value
Road Density Average Wage Index Total Revenue
Housing Starts

These key indicator constructs are nothing more than logical groupings of the key construction activity indicators identified in the preceding literature review. The following paragraphs are descriptions of these six proposed constructs. Population

Large increases in a county's total population drive the need for new buildings and infrastructure. Multiple urban centers will further drive the need for connecting transportation and distribution infrastructure. It is hypothesized that total population, population change and density within a county are positively correlated with the county's construction activity. The variables that will be used to measure this construct include (a) a county's total population and (b) population density. Geographic advantage

Secondary urban areas, or bedroom communities, typically develop around larger cities. This spillover effect contributes to the development of adjacent counties and is further amplified if the county is located between multiple large cities. It is hypothesized that a county's proximity to a larger urban center is positively correlated with the county's construction activity. The variable that will be used to measure a county's geographic advantage is a composite size/distance factor. This factor measures the relationship between the distance to an adjacent major city and the size of the adjacent major city.






28


To calculate this factor, the distances between the county seat of Florida's 10 most populated counties and the county seat each of the 67 counties were obtained. These distances were multiplied by the population of the 10 largest counties. This provided a value that is weighted both by distance and population. This value was then multiplied by negative one (-1) and divided by 1,000 to transform the order and scale. A county's final proximity factor is equal to the sum of the values of the two closest and most populated counties.

Initial infrastructure

Public construction projects tend to become larger and more complex as a county develops. Population growth drives the demand for residential housing, water resources, and energy delivery systems. Unpaved roads are paved, and existing road capacity is increased. Counties will also encourage the development of industrial parks and foreign trade zones in an effort to lure manufacturing and service employment. This initial construction activity is a predecessor to larger public construction projects such as toll highways, power plants, and water and wastewater treatment facilities. It is hypothesized that growth in a county's initial housing and infrastructure is positively correlated with a county's nonresidential construction activity. The variables that will be used to measure this construct include (a) the total daily vehicle miles traveled on a county's roadways,

(b) the total centerline miles of roadway, and (c) a county's road density.

Housing starts, or the total annual value of residential construction permits issued in a county, will only be used in the direct forecasting methodology which is discussed later in this chapter and Chapter 4.






29


Employment transition

As a county develops over the long-term, employment typically moves from

agriculture to manufacturing, then to the service sector. Shifts from the lower income employment sectors to the higher income sectors will stimulate investment in new facilities and infrastructure. This employment shift will also drive the need for a more highly educated work force. It is hypothesized that a shift from low to high-income employment sectors is positively correlated with a county's construction activity. The variables that will be used to measure a county's employment transition include (a) total personal income, (b) average wages, (c) and total construction payroll. Economic environment

A county's economic environment must be conducive to the growth of investment and employment. It is hypothesized that improvement of the economic environment within a county is positively correlated with the county's construction activity. The variables that will be used to measure a county's economic environment include (a) total employment, (b) gross sales, and (c) the Florida price level index. Financial resources

A county must have the financial resources to construct new facilities and

infrastructure. A county's bond rating is a reflection of its current financial position. Higher bond ratings enable a county to enter financial markets for essential borrowing at lower interest rates. It is hypothesized that an increase in a county's access to financial resources is positively correlated with a county's construction activity. The variables that will be used to measure a county's financial resources include (a) total tax collections, (b) taxable value of real property, and (c) a county's total revenue.






30


Summary of the Key Indicators

The various methodologies for segmenting the construction-market were reviewed. The approaches found in the literature review included segmentation by construction type, geographic location, and service type markets. They also included higher macro classifications such as private residential, private commercial (nonresidential), and public construction.

All of the market segments within the market place, and the market place itself, are shaped by changes in the marketing environment. These changes in the marketing environment are caused by the uncontrollable trends and forces that are external to the D&C organization. These environmental forces involve social, economic, technological, competitive, and regulatory changes. Our research has used the term key indicators to describe the forces and trends that are changing in the marketplace.

There appears to be a good deal of consistency in the literature regarding the key construction indicators at the macro level market segments of the industry (i.e., residential, commercial and public markets). But, which key indicators to use become more specific as a forecaster attempts to predict the activity within a major constructionmarket segment (e.g., housing, manufacturing, power, water supply, sewer/waste, industrial/petroleum, transportation, and telecommunications, environmental). An example of this is that a change in interest rates would have a greater effect on the housing market, than on the environmental construction-market.

Opportunities for our research to add to the existing key indicator literature include a comparison of the indicators over multiple markets, the application of the indicators at a specific geographic level (not just by construction type or service), and a deeper






31


understanding of the relationships between the key indicators and construction market activity.

A total of sixteen key construction indicators were selected from the literature review to be used in the research. To simplify the discussion regarding the key construction indicators, the indicators were grouped into six independent variable constructs. These key indicators will be used to forecast the public and private construction activity in the nonresidential construction-market at the Florida county level.

Approaches to Construction-Market Forecasting The final area of literature review for our research is a review of the approaches to construction-market forecasting.

Forecasting is the linchpin of business because it is attempting to reduce risk and uncertainty. It is central to setting targets, budgets, hourly rates for professional
services, planning of capital expenditure, and overhead recovery.40

The following sections of this chapter identify and describe the qualitative and quantitative forecasting techniques used during our research. Qualitative Techniques

The Modern Forecaster written by Levenbach & Cleary (1984) state that the

objective of qualitative forecasting techniques is to bring together in a logical, unbiased and systematic way all information and judgments that relate to the factors of interest.41 Levenbach & Cleary present five qualitative forecasting techniques including; Delphi






40 Fisher, N. (1986). (p. 123).

' Levenbach, H., & Cleary, J. P. (1984). The Modern Forecaster: The Forecasting Process through Data Analysis. Belmont, CA: Lifetime Learning Publications. (p. 15).






32


Method, Market Research Focus Groups, Panel Consensus, Visionary Technology Forecasts Using Curve Fitting, and Historical Analogue.42

Berkowitz et al. (2000) identify and describe six additional qualitative forecasting techniques including; Direct Forecasting, Lost-Horse Forecasting, Survey of Buyers Intentions, Sales force Survey, Jury of Executive Opinion, and Survey of Experts.43

Direct Forecasting. Direct forecasting is the simplest of the qualitative forecasting methods listed above. "Probably 99.9 percent of all sales forecasts are judgments of the person who must act on the results of the forecast - the individual decision maker."44 A direct forecast involves estimating the value of the forecast without any intervening steps. These estimates are opinions or judgments typically from experienced and competent executives inside a firm that know the market. They are gut feelings based on industry 45
conventional wisdom or learned experience. Direct forecasting techniques are commonly used to forecast something about which the amount, type, and quality of historical data are limited. Examples of everyday direct forecasts include; should we bid on the advertised construction project? How much money should we budget for the bid proposal and presentation? How much time should be allowed to drive to the meeting? Direct forecasting will be used as the only qualitative methodology in our research.

Although direct forecasting can be a quick and reasonably accurate prediction

methodology, most knowledgeable executives still want some level of mathematical or statistical analysis performed to validate the direct forecast. "The objective is to avoid 42 Levenbach, H., & Cleary, J. P. (1984). (pp. 15-17). 43 Berkowitz, E. N. et al. (2000). (pp. 248-250).

44 Berkowitz, E. N. et al. (2000). (p. 248).

45 Berkowitz, E. N. et al. (2000). (p. 248).






33


the use of only a single variable to represent a concept, and instead to use several variables as indicators, all representing differing facets of the concept to obtain a more well-rounded perspective."46 The historical data required for our research is adequately available to allow a quantitative analysis. The following sections of this chapter will identify and discuss the quantitative methodologies available for construction-market forecasting.

Quantitative Techniques

Levenbach & Cleary present eleven commonly used quantitative forecasting

techniques and further classifies them into either statistical (stochastic), or deterministic (causal) techniques.47 The seven statistical techniques include Summary Statistics, Moving Average, Exponential Smoothing, Box-Jenkins (ARIMA), TCSI Decomposition, Trend Projections, and Regression Model. The four deterministic techniques include Econometric Model, Intention to Buy (Anticipation Survey), Input-Output Model, and Leading Indicator. Montgomery (1976), Fisher (1986), Clapp (1987), Hair et al. (1998), and Berkowitz et al. (2000) all present similar quantitative techniques using somewhat different terminology.

Levenbach & Cleary outline six factors that should be considered before deciding on the most appropriate projection technique. These six factors include; (a) characteristics of the data, (b) minimum data requirements, (c) time horizon to be forecast, (d) accuracy desired, (e) applicability, and (f) computer and related costs.48 46 Hair, J. F. et al. (1998). Multivariate Data Analysis (5th ed.). Upper Saddle River, New Jersey: Prentice Hall. (p. 10).
4 Levenbach, H., & Cleary, J. P. (1984). (pp. 19-20). 4' Levenbach, H., & Cleary, J. P. (1984). (pp. 25-32).





34


Chambers et al. (1971) and Montgomery (1976) present similar selection factors using different terminology. Levenbach & Cleary go one-step further and developed a table to assist in the comparison and selection of the appropriate forecasting technique.49

The application of Levenbach & Cleary's table can quickly eliminate several

quantitative techniques for our research. A time horizon of two years eliminates all of the statistical techniques except summary statistics, trend projections, and regression models. It also eliminates two of the four deterministic techniques leaving the econometric model, and input-output models.

Brisbane Brown (1974), Roger Killingsworth (1990), and Standard & Poor's DRI (1999) have successfully applied econometric models to forecasting construction-markets and market costs. While there is widespread use of sophisticated econometric (causal) forecasts, there does not seem to be universal acceptance that econometric techniques produce consistently reliable and accurate forecasts (Armstrong, 1978; Granger and Newbold, 1977; Montgomery, 1976). Granger and Newbold contend that the econometric approach can be interpreted as a system in which a number of inputs are entered into a black box that transfers the values to an output. Montgomery states that the obvious limitation to the use of a causal model is the requirement that the independent variables must be known at the time the forecast is made. Another limitation of causal models is the large amount of computation and data compared with the time series model.




49Levenbach, H., & Cleary, J. P. (1984). (p. 28).

5 Levenbach, H., & Cleary, J. P. (1984). (p. 28).





35


Forecasting using a Summary Statistic technique was also evaluated for our

research. While summary statistics are a good tool for generating a profile of the overall data, this technique may be less accurate than a multiple regression statistical model. The two remaining techniques, regression and trend projections are the most applicable to our research.

Regression analysis

Multivariate Data Analysis written by Joseph Hair et al. (1998) presents a decision tree to assist with the classification and selection of the proper multivariate-regression technique.5' Hair et al. outline three judgments the researcher must make about the research objective and the nature of the data:52 (a) Can the variables be divided into independent and dependent classifications based on some theory? (b) If they can, how many variables are treated as dependent in a single analysis? (c) How are the variables, dependent and independent, measured (i.e., are they metric or non-metric)?

There are two types of multivariate statistical analysis techniques used in our

research, dependent and interdependent. These alternative regression techniques are used to estimate the level of construction activity in a county. These estimates are then compared to the results of the proposed momentum theory and to the actual known values. These dependent and interdependent regression techniques are discussed in the following paragraphs.







" Hair, J. F. et al. (1998). (pp. 20-2 1).
12 Hair, J. F. et al. (1998). (p. 18).






36


Dependent regression techniques

A dependent technique is one in which the prediction equation is dependent on, or estimated using, known dependent variable values. Our research involves one metric dependent variable (nonresidential construction activity) in a single relationship with multiple metric independent variables (key indicators). Therefore, the Hair et al. decision tree shows that either multiple regression or conjoint analysis could be used for this particular analysis.

Conjoint analysis is a multivariate technique used specifically to understand how respondents develop preferences for products and services.53 The most direct application is in new product or service development, allowing for the evaluation of complex products while maintaining a realistic decision context for the respondent.54 Understanding this intended use does not fit our research goals; conjoint analysis was not selected for use. Regression analyses, and more specifically linear and multivariateregression analysis, have been chosen as the most appropriate statistical techniques for our research.

Interdependent regression techniques

An interdependent technique is one in which the prediction equation is not

dependent on, or estimated using, known dependent variable values. It involves the simultaneous analysis of all the variables in the set. The goal is to find a way of consolidating the information contained in the original variables into an estimate or factor. This factor can then be objectively compared against the results of the momentum


5 Hair, J. F. et al. (1998). (p. 392).

5 Hair, J. F. et al. (1998). (p. 15).






37


forecasting methodology. Factor and cluster analysis are shown as the appropriate statistical techniques for our research in the Hair et al. decision tree.

The variables used in the factor and cluster analysis techniques cannot be classified as either dependent or independent variables. In these techniques, "all of the variables are analyzed simultaneously in an effort to find the underlying data structure of the entire set of variables."55 Factor analysis is the appropriate technique if the structure of the variables is to be analyzed. Cluster analysis is the appropriate technique if the cases (Florida counties) are to be grouped or classified. Trend analysis

John M. Clapp has made several significant literature contributions to real estate market analysis and forecasting. In his Handbookfor Real Estate Market Analysis,s6 Clapp presents a methodology for using regression for trend projection. Clapp's discussion also includes the limitations of the technique. For our research, a rolling one, two and three year trend projection analysis will be completed for the output variables of each of the forecasting methods. The purpose of these trend projections is to test the various methodologies for any inherent advantages relating to the duration of the forecast. The details of the trend forecast methodology used in our research are presented in Chapter 4.

Gap analysis

John Clapp has also made a number of contributions in the area of market and spatial gap analysis. "Market gap analysis determines whether there is (or will be) " Hair, J. F. et al. (1998). (p. 22).
56 Clapp, J. M. (1993). Dynamics of Office Markets. Washington, DC: The Urban Institute Press. (pp. 226227).





38


unsatisfied demand in the entire market area."57 Clapp outlines three methods for measuring market gaps: (a) gravity gap analysis, (b) expenditure-sales gap analysis, and

(c) comparison of projected supply and projected demand. The first approach estimates the demand at a specific location. The last two approaches are used to determine the differences between supply and demand over a larger market area. The difference between the last two methods is that the second method uses the actual number of units for the supply and demand projection, whereas the last method uses the percentage change. Because the data for the actual number of units is available, an adaptation of the second methodology (expenditure-sales gap analysis) will be used in our research.

Our research uses gap analysis for several purposes. First, gap analysis will be

used to compare a county's construction demand to its supply of construction resources. This will identify the Florida counties with construction demand that is larger or growing more rapidly, than the available supply. Second, gap analysis will be used as a tool to group the counties by their positive gap, balanced gap, or negative gap. A positive gap indicates that additional construction resources are needed in the county to meet the construction activity demand. Balanced gap indicates a balance between the supply of construction resources and the demand of construction activity. A negative gap indicates a surplus of construction resources exist for the corresponding demand for construction activity. Finally, gap analysis will be used to predict, and rank order, the level of construction activity in a county. The details of the different methodologies for applying gap analysis are presented in Chapter 4.


7 Clapp, J. M. (1987). (p. 179).





39


Law of Universal Gravitation

In the 1680's, Sir Isaac Newton developed one of the most influential theories in Physics, the Law of Universal Gravitation. In 1929, William J. Reilly applied the logic of the law of universal gravitation to trade area analysis. Reilly's gravitational forecasting technique estimated the point between two cities where a customer would be equally likely to travel to one or the other.58 Huff (1964) applied Reilly's gravitational model to estimating the visitation rates of customers from a given neighborhood to a given store.59 Clapp (1987) improved and extended Reilly's work with the concepts of Gravity Capture Analysis,60 and Gravity Gap Analysis.61 Momentum analysis

Sir Isaac Newton also developed the three laws of motion. These laws of motion were published in his book Principia and remain as the cornerstone laws in the natural and physical sciences. Included in these three laws is the theory of linear momentum. The theory of linear momentum is used to mathematically relate the size and speed of an object or system.

The logic of Sir Isaac Newton's momentum theory has been successfully applied in various areas of the social sciences and the natural sciences. Examples include; population momentum (World Bank Group 2003),63 political campaign momentum


5 Reilly, J. W. (1959). Methodsfor the Study of Retail Relationships. Austin, TX: University of Texas. 59 Huff, D. L. (1964, July). Defining and Estimating a Trade Area. Journal of Marketing. 60 Clapp, J. M. (1987). (pp. 173-178).
61 Clapp, J. M. (1987). (pp. 180-186).
62 Newton, Isaac. (1687). Principia, Translated by Andrew Motte 1729. 63 Development Education Program Web, Glossary. The World Bank Group. Retrieved October 08, 2003, from http://www.worldbank.org/depweb/english/modules/glossary.htmi.






40


research (Momentum Analysis Opinion Research, LLC 2003),64 stock market momentum forecasting (Martin Pring 1993,65 Hugh Clark 200266), and economic momentum measurement (Missouri Economic Research and Information Center 2003).67 These last two applications are of the most interest to our research and are expanded upon below. Momentum forecasting in the stock market

In the 1970's Dr. George Lane developed a securities market forecasting indicator known as the stochastic oscillator, or momentum indicator. Over the last 25 years Lane's forecasting methodology has been used and adapted by the financial industry to measure a securities rate of change. The Keystone Commodity Trading Guide (2001)68 lists several type of common momentum indicators in use including: The Stochastic Oscillator, Rate of Change, Smoothed Rate of Change, Momentum Index, RSI, Williams %R, Commodity Channel Index, and the Moving Average Convergence/Divergence method. Other published methods include the Ford Value/Momentum model,69







4 Momentum Analysis. Momentum Analysis Opinion Research, LLC. Retrieved October 08, 2003, from http://www.ballot.org/resources/MomentumAnalysisBrochure.doc. 65 Pring, M. (1993). Martin Pring on Market Momentum. Sarasota, FL: International Institute for Economic Research, Inc. (p. 2); and Pring, M. (2002). Momentum Explained. New York: McGraw-Hill.

6 Clark, H. (2002). Smart Momentum, the Future of Predictive Analysis in the Financial Markets. Chichester: John Wiley & Sons, LTD. (p. 7).
67 Economic Indicators. Missouri Economic Research and Information Center. Retrieved October 13, 2003, from http://www.ded.state.mo.us/business/researchandplanning/indicators/momentun/index.shtffd. 61 Momentum Indicators. Keystone Commodity Trading Guide. Retrieved October 13, 2003, from http://www.keystone-web.com/technicals/momentum.htmt. 69 Value/Momentum Sector Analysis - August 31, 2001. Ford Equity Research. Retrieved October 13, 2003, from http://www.fordequity.com/htmil/documents/SsO8Ol.pdf.






41


Trendcast System,70 Stochastic Momentum Index,71 Relative Momentum Index,72 and the CBR Stock Market Momentum Indicator.73

While there are many differences between these methodologies, these momentum indicators generally measure the price of a security relative to the high/low range over a set period of time. Momentum investing is based on the idea that stocks which have performed well over some interval in the past will tend to perform well in the future.74 This same logic is applied to a county's level of construction activity over time in our research.

Economic momentum

The Missouri Economic Research and Information Center (MERIC) publishes an annual Index of State and Economic Momentum. This index, developed by the late State Policy Reports editor Hal Hovey, is a composite of percentage changes in personal income, population, and employment at the county level. The index measures momentum in a county relative to the overall economic momentum of the state. "An index equal to 0 means the county realized average economic growth during the decade.






70 Answers to Frequently Asked Questions. Trendcast, LLC. Retrieved October 13, 2003, from http://www.trendcast.com/amazing/faq.htm.
7 Stochastic Momentum Index. Paritech Inc. Retrieved October 13, 2003, from http://www.paritech.com/education/technical/indicators/momentum/stochastic 1.asp. 72 TradingSolutions Function Library, Relative Momentum Index [RMIJ. Trading Solutions. Retrieved October 13, 2003, from http://www.tradingsolutions.com/functions/RelativeMomentumlndex.html.
7 Stock Market Momentum Indicator. Commodity Research Bureau. Retrieved October 13, 2003, from http://www.crbtrader.com/crbindex/nsmi.asp.
7 Capeci, J. D. & Campillo, M.. (April 2002). Global Sector Momentum in the Emerging Markets. Cambridge, MA: Arrow Street Capital. (p. 2).






42


An index less than zero indicate relatively sluggish growth, while an index greater than zero indicates relatively prosperous growth."75

Hovey improved the original economic momentum index by adding the measures of county economic share and influence. MERIC defines economic share as the percentage of the state's economy that is accounted for by an individual county. The economic share is measured as the average of the percentage of the state's employment, population, and personal income that occurs in a particular county. MERIC defines economic influence as the product of the momentum index and economic share score, and is calculated by multiplying the two indices. Thus, a county with a high momentum score and a large economic share has a large level of economic influence, and is considered an important economic driver for the state.

MERIC has also applied this same methodology to the regional and state

geographic level. The methodology for measuring economic momentum is the same for these geographic levels.

While these gravitational and momentum methods may seem unrelated to the

construction-market forecasting focus of our research, the significance of Reilly, Huff, Clapp, Lane and Hovey's research is that they have demonstrated how the framework of natural science theories can be successfully adapted and applied to social science research.

Summary of Forecasting Approaches

No single model can be considered universally adaptable to any given forecasting situation. Thus a basic principal is to utilize more than one projection technique. The


75 Economic Indicators. Missouri Economic Research and Information Center. (2003).






43


purpose of using more than one methodology is to insure that the forecasting approach will be as flexible as possible and that the forecaster's judgment is not overly dependent on one projection technique.76

Opportunities for our research to add to the existing literature include the

development of an accurate forecasting technique that is more understandable than complex regression approaches but more meaningful than simple direct approaches. Another opportunity includes developing a technique that integrates the variable dimensions of size, rate of change, and influence. Finally, the literature appears to be missing the application and comparison of multiple forecasting methodologies over several diverse markets.

Our research uses six different techniques identified in the previous literature

review to forecast a county's construction activity. These six methodologies include; 1), direct forecasting, 2) multivariate-regression, 3) factor analysis, 4) gap analysis, 5) MERIC economic momentum analysis and 6) momentum analysis. The following chapter will discuss how momentum theory was derived and apply it to strategic construction-market forecasting. In Chapter 4, the estimates from all six of these methodologies will be compared against the actual values of nonresidential construction activity for all of the counties in the state of Florida.

Summary of Research Questions

* Which, and what type of, key indicators best predict a county's nonresidential
construction activity?

* What are the relationships between these key indicators and county construction
activity?


76 Levenbach, Hans & Cleary, James P. (p. 35).





44


" Can a more accurate method of forecasting construction activity be developed and
applied such as the proposed momentum methodology?

" Are complicated forecasting methods really more accurate than simple, more direct,
approaches?

* Can the various construction-markets be segregated using methodologies such as
gap or cluster analysis so the highest potential markets can be prioritized and
pursued?

* Where are the best new markets for a large D&C firm in the state of Florida?













CHAPTER 3
MOMENTUM THEORY DERIVATION AND APPLICATION

Make things as simple as possible, but not simpler.I

-Albert Einstein responding to the question of how he explained the difficult theory of relativity with such a brief equation as E=MC2 The philosophy behind Einstein's statement was used as the premise behind the development of the momentum forecasting approach. As discussed in Chapter 2, the momentum analysis approach was designed to be less complex but more meaningful than traditional statistical regression approaches.

This chapter discusses all of the steps taken during our research to derive and apply the methodology of momentum theory to strategic construction-market forecasting. The chapter begins with an introduction to the momentum forecasting theory. The differences between this momentum methodology and multivariate statistical regression approaches are discussed. This is followed by a brief overview of the data sources and sample. Finally, the procedure for applying and analyzing county momentum is then detailed, and a momentum index is derived.

In the next chapter of our research, the results of this momentum analysis are then comparatively validated against five alternative forecasting methods.

Momentum Theory Introduction

Momentum is a commonly used term to describe objects in motion. The concept of momentum is used to relate the size and speed of an object or system. Examples include; ' Brallier, J. (2002). Who Was Albert Einstein? New York, Grosset & Dunlap. (p. 47).


45






46


a fast train has more momentum than a slow train; the football team has lost its momentum going into the fourth quarter; or the latest government reports show that the economy may be gaining momentum.

As discussed in Chapter 2, the theory of linear momentum was derived from Sir Isaac Newton's Three Laws of Motion. The linear momentum of an object is defined as a product of its mass and its velocity.2 This momentum theory is mathematically shown in Equation 3-1. The magnitude of momentum "p" at any moment is equal to the numerical product of the mass "m" times the velocity "v".

(Eq. 3-1) p = mv

The importance of the theory of momentum for our research is that it is a conserved quantity. The Law of Conservation of Momentum is stated as follows;

The total momentum of an isolated system of bodies remains constant.3


Although the momentum of each of the objects changes as a result of the collision, the sum of their momentum is found to be the same before and after the collision.

A system is defined as a set of objects that interact with each other. The total

momentum of a system "Ptota1" is equal to the sum of the momentum in the components "pn" of the system. This conservation of momentum theory is mathematically shown in Equation 3-2.

(Eq. 3-2) Ptotai =P1+P2... + or Ptotal = mIvI + m2v2... + movn

Newton's theory assumed that the external forces (i.e., friction) on the isolated

system of objects under study are zero. In our research, different key indicators (objects)

2 Giancoli, D. C. (1980). Physics. Englewood Cliffs, NJ: Prentice-Hall, Inc. (p. 116).

3 Giancoli, D. C. (1980). (p. 120).






47


will influence a county's construction activity at different levels. For example, it is likely that a county's construction activity is more influenced by a change in its total population than by a change in the education level of its population. For this reason, the influence variable " i " will be added to the calculation of momentum in a county. The final equation used to calculate the total momentum in a county is shown in Equation 3-3.

(Eq. 3-3) Ptota = P I + P 2... + Pn or Ptotai = mivi i + m2v2 i2... + mnvn in

Differentiation between Momentum and Regression Approaches

There are two primary differences between the proposed momentum model and traditional regression approaches. The following paragraphs are a summary of the two most significant differences including; (a) variable dimensions, and (b) multivariate measurement.

First, traditional multivariate-regression analysis equations are constructed with a linear combination of one-dimensional variables with empirically determined weights. The variables are specified by the researcher, and the weights are determined by the regression technique. The result is a single output value representing a combination of the entire set of variables that best achieves the objective of the specific multivariateregression analysis. A typical regression variate can be stated mathematically as shown in Equation 3-4.

(Eq. 3-4) Variate output value (Y)= fo + f31XI + 02X2 + 03X3 + ... + OnX.

The momentum analysis equation is also constructed with a linear combination of variables that are specified by the researcher. The difference in the momentum approach is that each variable includes three dimensions. These dimensions include the relative size, rate of change, and empirically determined influence of the variable. These three dimensions are combined into one new variable value (i.e., momentum). This new






48


variable value provides a more meaningful representation of the original variable and the relationships within, and between, other variables. The momentum variate can be stated mathematically as shown in Equation 3-5.

(Eq. 3-5) Variate output value (Ptotal) m1v1 i i + m2v2 i2... + mnvn in or
PtotalPI+P2...+Pn

The second fundamental difference between the techniques is that the output of the momentum equation is really a multivariate measurement or summated scale, not an estimate of the true dependent variable value. As with regression analysis, the result of the momentum equation is a single value representing a combination of the entire set of variables. But in the case of momentum, the variate value is derived without using statistical regression. The momentum variate value is a summation of the variables joined together as a composite measure of the concept under study (i.e., construction activity). These two different variate measurement techniques are reconciled in the research when the estimated variate values are calculated for each case, and the cases (counties) are rank ordered and compared.

Data Sample and Data Sources

Historical data was collected for all of Florida's 67 counties for use in our research. A map of the State of Florida's 67 counties is shown in Figure 3-1. A thirteen year time period was used for our study that started in 1990 and ended in 2002. Our research involves the analysis of one single dependent variable and its relationships to sixteen independent variables. This resulted in a data set of over 14,800 values (67 counties x 13 years x 17 variables). The data for our research was collected from a variety of secondary sources. Missing or erroneous data was corrected by using mean or trend






49


projection techniques. Table 3-1 outlines each of the variables including the variable name, type, unit, description, and source.

The actual variable data for each of the thirteen years in the study is included in

Appendix B. Due to data license agreement restrictions, the nonresidential permit values for the years 1996 through 2002 are not included. Other data required to replicate our research is available by contacting the author.

Momentum Theory Applied to Strategic Construction-Market Forecasting

Applying the logic of momentum theory to strategic construction-market

forecasting is fairly straightforward. The momentum theory variables introduced above have been redefined for the purposes of our research and are shown in Table 3-2.

The measurement of momentum in each county is relative to the other 67 counties in Florida. A county with high construction momentum (p) would be characterized as being a large existing market (m), with rapid positive change (v) in the key variables that most influence (i) construction activity. Conversely, a county with low construction momentum (p) would be characterized as being a small existing market (m), with slow or negative change (v) in the key variables that most influence (i) construction activity. Methodology for Analyzing County Momentum

The following sections of this chapter detail the methodology for calculating a county's momentum and the corresponding momentum index and slope. The seven general steps for the momentum analysis are outlined in Figure 3-2.









50


Figure 3-1. Map of the 67 counties in the State of Florida.


Waften~~Hmito Wahitn Gddn Jf Nassau



Lberry Wakulla Taylor Swne olmtsa


Gu Franklin LafBrydfor Clay San


Mii i iAlchuFlagler
Levy


Volusia


citrus Lake Sumntr nola0 HernandoOrange

Paso


HAsboroughOsel Pinellas Pl rvr

Indian River

an..Ha Okeechobee

- Higha St. Lucie
San.s., Desolo


Charlotte Gld Lee Harl Patim Beach




Cc. Ir Broward




Mionro








51


Table 3-1. Research variable data, descriptions, and sources.
No. Variable Variable Name Type Unit Description Source
Code


Dependent Variable: I NRPERMIT Nonresidential
Construction Permits



Independent Variables: I POP Total
Population


Metric USD$ Total annual value of
Ratio (1,000's) nonresidential construction
permits issued per county.


1) Building Permit Activity in Florida 1990 to 2003, University ofFlorida Bureau of Economic and Business Research. 2) Annual Construction Starts Data 1996-2002. McGrawHill Construction Dodge.


Metric Person The annual computed number Florida Statistical Abstracts 1990 to 2003,
Ratio each of persons living in a county University ofFlorida Bureau ofEconomic and
(1,000's) (resident population). Business Research.


Population Metric Persons The annual computed number Florida Statistical Abstracts 1990 to 2003,
Density Ratio (1,000's) of persons living in a county University ofFlorida Bureau of Economic and
per Sq. divided by the total area Business Research. Mile (square miles) of the county.
Proximity-Size Metric Factor Annual factor measuring the 1) Mileage Between Florida Cities, Intercity Factor Interval relationship of proximity and Mileage Spreadsheet (Rev. 12/16/03). Florida
population between counties. Department of Transportation. 2) Florida Statistical Abstracts 1990 to 2003, University of Florida Bureau of Economic and Business Research.
Daily Vehicle Metric Mile An annual measure of the Public Road Mileage and Miles Traveled,
Miles Traveled Ratio total traffic on a county's 1990-2002. Florida Department of
roads. It is the product of the Transportation. average daily traffic count and
the length of the road.
Centerline Metric Mile The annual existing total Public Road Mileage and Miles Traveled,
Miles of Ratio length of a county's roads. 1990-2002. Florida Department of


Roadway Road Density


Metric Miles
Ratio per
square
mile


7 TOTPINC Total Personal Metric USD$
Income Ratio (million
s)
8 AVEWAGE Average Wage Metric USD$
Ratio


Transportation.
The annual total length of a 1) Public Road Mileage and Miles Traveled, county's roads divided by the 1990-2002. Florida Department of total area (square miles) of the Transportation. 2) Florida Statistical Abstracts county. 2002, University of Florida Bureau of
Economic and Business Research.
Total annual amount of Florida Statistical Abstracts 1990 to 2003,
personal income on a place of University ofFlorida Bureau ofEconomic and residence basis in a county. Business Research. Average annual earnings per Florida Statistical Abstracts 1990 to 2003, job in a county. University of Florida Bureau of Economic and
Business Research.


9 CPAYROLL Construction Metric USD$ Average monthly private Florida Statistical Abstracts 1990 to 2003,
Payroll Ratio (1,000's) construction payroll covered University ofFlorida Bureau ofEconomic and
by unemployment Business Research.
compensation law in a county.
10 PLINDEX Price Level Metric Factor Annual Price Level Index of Florida Statistical Abstracts 1990 to 2003,
Index Interval prices of major items in a University ofFlorida Bureau of Economic and
county. Business Research.
It GSALES Gross Sales Metric USD$ Gross sales reported to the Florida Statistical Abstracts 1990 to 2003,
Ratio (million' Florida Department of University ofFlorida Bureau ofEconomic and
s) Revenue. Business Research.
12 TOTEMPLY Total Metric Person Total annual number of wage Florida Statistical Abstracts 1990 to 2003,
Employment Ratio each and salary jobs in a county. University ofFlorida Bureau ofEconomic and
Business Research.
13 TOTREV Total Revenue Metric USD$ Total annual revenue by or Florida Statistical Abstracts 1990 to 2003,
Ratio (1,000's) within a county University ofFlorida Bureau ofEconomic and
Business Research.
14 TOTTAX Total Taxes Metric USD$ Total annual tax collections Florida Statistical Abstracts 1990 to 2003,
Ratio (1,000's) by or within a county University ofFlorida Bureau ofEconomic and
Business Research.
15 ALVALUE Assessed Land Metric USD$ Total annual assessed value of Florida Statistical Abstracts 1990 to 2003,
Value Ratio (Million commercial land in a county. University ofFlorida Bureau ofEconomic and
(Commercial) s) Business Research.
16 RPERMIT Residential Metric USD$ Total annual value of Building Permit Activity in Florida 1990 to
Permits Ratio (1,000's) residential construction 2003, University ofFlorida Bureau of
permits issued per county. Economic and Business Research.


2 PDENSITY



3 PROX 4 DVMT


5 CLMILES


6 RDENSITY






52


Table 3-2. Comparison of momentum theory variable definitions. Variable Newton's momentum theory definition. Construction-market momentum definition.
Ptotal Total momentum of a system. Total momentum in a
construction-market (county).

p Momentum of a component of the system. Momentum of a key variable.

m Mass of a component of the system. Mass (size) of a key variable.

v Velocity of a component of the system. Velocity (percent change) of a
key variable.

i Analogous to an external force on an Influence (strength) of a key
object or isolated system. variable.

1 2 3 4 5 6 7
Calculate Calculate Calculate Calculate Calc total Calculate Calculate variable variable variable variable county momentum momentum
mass velocity j influence momentum momentum index index slope
v "I"t indexe" "oslope"

Figure 3-2. Seven general steps of momentum analysis.


Mass of key indicators

The first step is to calculate the mass "in" of the key variables. Mass is calculated using the actual value of the key variable. The variable value is normalized and is expressed in terms of its proportion to the range of values for the same variable. A value of one (1) has been added to the equation to eliminate a "0" effect during multiplication in the final momentum calculation. The range of output values for this calculation is 1.0 to 2.0. This mass calculation is shown in Equation 3-6.

(Eq. 3-6) M= X-_Xi. + 1
Xmax - Xmin

In Equation 3-6, "in" equals the calculated mass of the variable, Xn equals the

variable's actual value, Xmin equals the minimum actual value of the same variables, and Xmax equals the maximum actual value of the same variables.






53


Example mass calculation: If the variable has minimum and maximum values of 200 to 300 respectively, and a particular county has a value of 240, the mass for that variable would equal 1.40.

Velocity of key indicators

The next step is to calculate the velocity "v" of the key variables. The variable value will be expressed in terms similar to the previous mass (in) calculation. The only difference is that velocity is measuring the annual percent change of the variable, not its size. The velocity calculation is shown in Equation 3-7.

The variable value is normalized and is expressed in terms of its proportion to the range of values for the same variable. A value of one (1) has been added to Equation 3-7 to eliminate a "0" effect during multiplication in the final momentum calculation. The range of output values for this calculation is 1.0 to 2.0.

(Eq. 3-7) v = X i + 1
Xmax - Xmin

In Equation 3-7, "v" equals the calculated velocity of the variable, Xn equals the variables annual percent change, Xmin equals the minimum annual percent change of the same variables, and Xmax equals the maximum annual percent change of the same variables. The annual percent change of a variable was calculated using the example Equation 3-8.

(Eq. 3-8) Annual percent change = (Year 2001 - Year 2000 / Year 2000)

Example velocity calculation: If the variable has minimum and maximum values of 2.0% to 3.0% respectively, and a particular county has a value of 2.4%, the velocity for that variable would equal 1.40.





54


Influence of key indicators

The next step is to calculate the influence (i) of a key variable on a county's total construction activity. The Pearson Correlation Coefficient is a statistical measure that has been used as the value for influence in our research. The Pearson correlation coefficient is a measure of linear association between two variables. Values of the correlation coefficient range from -I to 1. The sign of the coefficient indicates the direction of the relationship, and its absolute value indicates the strength, with larger absolute values indicating stronger relationships. The statistical software Statistical Package for the Social Sciences (SPSS) Base version 11.5 (2002) was used to calculate the Pearson correlation coefficient on the dataset. This influence calculation is shown in Equation 3-9. A value of one (1) has been added to the equation to eliminate a "0" effect during multiplication in the final momentum calculation. The range of output values for this calculation is 1.0 to 2.0.

(Eq. 3-9) i = |+I

In Equation 3-9, " i " equals the calculated influence of the variable, and "lu "is the absolute value of the Pearson correlation coefficient.

The Pearson coefficient was computed for all sixteen independent variables

included in the data sample during the years 1990 through 2002. A two-tailed test of significance was used. In comparison to the other variables, the AVEWAGE, PLINDEX, PDENSITY, RDENSITY and PROX independent variables consistently demonstrated weaker correlation (< 0.750) to the dependent variable nonresidential permits (NRPERMIT). These five independent variables were dropped and are not included in the analysis of county momentum or the other forecasting methods.





55


In addition to these five excluded variables, the independent variable RPERMIT will be used in our research only as a variable in the directforecast methodology. The RPERMIT variable will be excluded from the momentum analysis and all other methods in our research. Out of the sixteen variables originally proposed in Chapter 2, the ten remaining variables will be used as the key indicators for all of the forecasting techniques in our research.

The results for the Pearson's correlation calculation are presented in Table 3-3

along with their minimum, maximum and average correlation. The variables are sorted by their average correlation in descending order. The five excluded variables have been shown below the dashed line for reference purposes. A detailed discussion of each variable's correlation to NRPERMIT is included in Chapter 5. Momentum of key indicators

The next step is to calculate the momentum (p) for each key variable. This momentum calculation is shown in Equation 3-10. The momentum for a particular variable (pa) is the product of the key variable's mass (m), velocity (v) and influence (i). The maximum momentum value for a variable equals 8 (2 x 2 x 2), and the minimum momentum value for a variable equals 1 (1 x 1 x 1).

(Eq. 3-10) pn = mnvn in

A variable's momentum can also be thought of as a three-dimensional or

volumetric measurement. This three-dimensional relationship is shown in Figure 3-3(a). Total momentum of a county

The final step is to calculate the total momentum (Ptosta) in a county. As discussed earlier in this chapter, total momentum is a conserved quantity that is equal to the sum of the momentum in the components (variables) of the system (county). This relationship is






56


shown in Figure 3-3(b). Using Equation 3-2, the total momentum in a particular county is the sum of the momentum for each key variable within the county. The results of the momentum calculation for each year of the study are shown in Table 3-4. The 67 counties were sorted by their average momentum in descending order.




vii



(a) Variable momentum (p,) (b) Total momentum (Ptotal)


Figure 3-3. Three-dimensional relationships of momentum. (a) Variable momentum, (b)
Total momentum.


Derivation of the Momentum Index

The construction-market momentum index is a measure of the cumulative value of a county's total momentum. The base year for the index is 1990 with an initial value of 100. The calculated value of a county's total momentum was divided by 10 for scale.

An example equation for calculating the momentum index (OINDEX) for a county in a given year is shown in Equation 3-11.

(Eq. 3-11) OINDEX1991 = P1990 + (P1991 / 10)

For example; if a county's 1990 momentum index value was 100, and its 1991

momentum value was 35, the county's 1991 momentum index value is 103.5 (i.e., 100 + (35/10) = 103.5).










Table 3-3. Pearson Correlation Coefficient to nonresidential permit activity (NRPERMIT).


Year
VARIABLESa 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002


NRPERMIT 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
TOTTAX 0.846 0.976 0.963 0.970 0.973 0.976 0.958 0.982 0.977 0.974 0.973 0.961 0.976
ALVALUE 0.841 0.981 0.975 0.958 0.976 0.974 0.970 0.973 0.977 0.960 0.969 0.953 0.969
TOTEMPLY 0.784 0.969 0.963 0.965 0.965 0.949 0.968 0.973 0.959 0.951 0.947 0.954 0.973
GSALES 0.763 0.960 0.956 0.954 0.961 0.940 0.972 0.967 0.976 0.946 0.936 0.938 0.966
CPAYROLL 0.872 0.957 0.933 0.960 0.948 0.953 0.922 0.962 0.934 0.960 0.939 0.914 0.960
DVMT 0.791 0.940 0.928 0.968 0.940 0.938 0.941 0.951 0.949 0.940 0.920 0.937 0.979
POP 0.759 0.942 0.947 0.958 0.946 0.920 0.951 0.937 0.949 0.905 0.900 0.918 0.977
TOTPINC 0.819 0.928 0.916 0.931 0.918 0.910 0.913 0.922 0.920 0.899 0.897 0.899 0.983
TOTREV 0.625 0.919 0.961 0.891 0.903 0.835 0.951 0.863 0.885 0.831 0.842 0.883 0.940
RPERMIT' 0.840 0.916 0.915 0.845 0.836 0.918 0.730 0.880 0.854 0.892 0.811 0.821 0.848
CLMILES 0.720 0.830 0.830 0.905 0.855 0.828 0.830 0.846 0.803 0.805 0.807 0.844 0.868
AVEWAGE 0.639 0.673 0.608 0.629 0.687 0.725 0.699 0.732 0.662 0.702 0.716 0.692 0.729
PLINDEX 0.536 0.609 0.640 0.627 0.590 0.625 0.655 0.674 0.624 0.701 0.688 0.630 0.681
PDENSITY 0.611 0.596 0.587 0.628 0.608 0.635 0.581 0.657 0.624 0.651 0.628 0.632 0.675
RDENSITY 0.501 0.432 0.428 0.500 0.443 0.468 0.402 0.481 0.418 0.470 0.451 0.465 0.496
PROX 0.392 0.331 0.339 0.368 0.352 0.354 0.328 0.365 0.343 0.375 0.370 0.389 0.370
Maximum 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
Minimum 0.392 0.331 0.339 0.368 0.352 0.354 0.328 0.365 0.343 0.375 0.370 0.389 0.370
Average 0.739 0.830 0.826 0.833 0.824 0.829 0.817 0.841 0.824 0.830 0.820 0.822 0.853


a Variables are sorted by average correlation. b RPERMIT is used as a variable in the Direct forecasting methodology only.


Maximum Minimum
1.000 1.000
0.982 0.846
0.981 0.841
0.973 0.784
0.976 0.763
0.962 0.872
0.979 0.791
0.977 0.759
0.983 0.819
0.961 0.625
0.918 0.730
0.905 0.720
0.732 0.608
0.701 0.536
0.675 0.581
0.501 0.402
0.392 0.328


Average
1.000 0.962 0.960
0.948 0.941 0.940 0.932
0.924 0.912
0.856
0.854 0.829 0.660 0.637
0.624 0.458 0.360






58


The momentum index for each of the 67 counties was calculated and plotted from 1991 through 2002. A rolling one, two and three year trend projection analysis was completed for the momentum index. This trend projection analysis methodology is described in Chapter 4. The results of the momentum index calculations are included in Table 3-5. The 67 counties were sorted by the 2002 momentum index in descending order. This momentum index will be used as an indicator to forecast a county's nonresidential construction activity (NRPERMIT). Momentum Index Slope

The results from the preceding momentum index calculations were plotted and are shown in Figure 3-4. This graph shows the momentum index for all 67 counties from 1991 through 2002. When observing this graph it can be seen that the lines of the counties with the highest momentum indexes have a greater line slope while the lines of the counties with the lowest momentum indexes have a lower slope. This slope will be used as the second indicator from this momentum methodology to forecast a county's nonresidential construction activity.

The slope of a county's index in a specific year was measured by the slope of the linear regression line through the momentum index data points for the preceding five years. The slope is the vertical distance divided by the horizontal distance between any two points on the line, which is the rate of change along the regression line. The momentum index slope for each of the 67 counties was calculated. A rolling one, two and three year trend projection analysis was completed for the momentum slope. This trend projection analysis methodology is described in Chapter 4. The results of the momentum index slope calculation are discussed in Chapter 6.






59


Summary of Momentum Theory

The intent of this chapter was to introduce a new forecasting methodology based on Sir Isaac Newton's natural science theory of momentum. This unique new methodology was applied to strategic construction-market forecasting.

The two most significant differences between this new momentum methodology and traditional statistical regression approaches were discussed and include; (a) variable dimensions, and (b) multivariate measurement.

The data sample and data sources used in our research were reviewed. Finally, the procedure for applying and analyzing county momentum was detailed. This procedure included; calculating the mass, velocity and influence of the key indicators; calculating the key indicator momentum and total county momentum; and calculating the momentum index and its corresponding slope.

In the next chapter of our research, the results of this momentum analysis are

comparatively validated against five alternative forecasting methods. These alternative methodologies were previously outlined in Chapter 2.









60




Table 3-4. Total momentum of a county for years 1991 through 2002.
County' 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Average
Dade 52.84 50.73 57.08 52.09 50.49 54.46 57.41 57.59 57.54 57.56 56.82 53.38 54.83
Broward 45.45 45.80 4734 47.14 43.33 47.84 50.24 49.07 50.99 51.22 50.79 47.87 48.09
Orange 41.85 40.37 41.50 37.98 40.00 45.25 47.51 46.24 47.11 46.59 44.75 43.24 43.53
Palm Beach 40.76 40.31 42.17 40.46 39.15 43.73 45.36 45.42 44.94 46.62 44.41 43.63 43.08
Hillsborough 39.77 40.03 39.47 40.52 38.20 43.54 45.29 45.85 46.87 43.29 43.83 44.43 42.59
Duval 37.87 37.35 3961 37.08 37.49 40.54 44.54 41.01 41.19 40.12 40.15 39.45 39.70
Pinellas 37.62 36.44 3736 36.72 34.82 38.65 41.19 39.53 39.50 40.45 38.99 37.87 38.26
Lee 34.03 31.62 34.76 34.28 32.34 34.34 37.25 35.82 37.27 35.92 35.96 36.77 35.03
Polk 32.75 31.63 31.80 32.66 29.47 33.65 36.29 34.61 35.28 35.72 32.26 34.12 3335
Brevard 34.10 33.53 32.25 31.77 29.17 33.45 35.25 32.52 34.75 35.67 33.35 32.43 33.19
Seminole 31.45 31.53 32.06 29.96 29.43 33.12 36.20 33.65 36.43 32.97 32.64 34.57 33.00
Collier 32.41 30.40 3332 30.94 29.76 32.81 35.62 35.18 34.76 33.74 33.78 31.60 32.86
Volusia 33.01 31.01 31.83 30.92 29.91 32.64 34.52 31.92 34.14 33.68 32.23 32.19 3233
Sarasota 31.15 30.45 30.84 31.45 28.75 32.36 35.12 32.98 33.58 32.62 33.24 32.29 32.07
Marion 30.30 30.99 30.83 31.51 28.51 31.66 34.78 33.63 34.40 31.07 30.65 31.73 31.67
Manatee 31.77 29.94 30.11 31.49 29.22 30.72 32.90 30.27 33.27 31.49 31.13 31.96 31.19
Osceola 29.53 31.17 30.73 29.52 28.03 31.32 31.84 34.06 33.96 30.54 30.06 32.37 31.10
Pasco 32.36 29.30 29.55 29.39 27.54 31.85 33.37 30.82 33.46 30.59 31.54 32.54 31.03
Lake 29.44 30.06 30.27 29.83 28.37 31.11 33.64 31.79 33.66 30.66 30.71 30.71 30.85
Escambia 29.98 30.86 30.45 30.37 27.74 31.92 33.05 31.00 3236 30.32 29.72 31.04 30.73
Saint Jolms 29.43 29.31 29.63 29.69 28.89 29.54 34.61 31.84 32.34 31.01 29.92 31.94 30.68
Leon 30.89 29.98 30.14 30.33 28.42 31.14 31.74 31.21 32.39 29.39 29.65 30.72 30.50
Walton 30.32 29.56 2834 28.43 28.13 30.90 32.36 31.22 32.18 29.47 30.41 31.29 30.23
Alachua 29.66 30.14 29.74 29.49 27.81 30.72 31.47 30.84 31.53 29.42 30.86 29.85 30.13
Okaloosa 31.14 29.45 31.03 28.99 29.38 30.36 30.29 31.93 30.46 29.39 29.37 29.22 30.10
Santa Rosa 31.30 31.53 2935 29.68 27.42 31.10 32.15 30.33 28.92 29.37 29.86 29.30 30.03
Hernando 28.49 29.97 2926 31.39 27.27 29.55 33.59 29.99 29.52 29.03 29.34 30.15 29.80
Flagler 29.21 27.87 30.88 30.48 28.92 30.15 31.32 29.68 31.78 29.30 28.95 28.81 29.78
CIarlotte 31.16 28.33 29.07 28.66 26.81 29.67 31.37 30.44 30.97 29.61 30.72 30.08 29.74
Saint Lucie 31.36 28.24 2852 29.55 27.84 2930 30.39 29.70 30.69 30.08 29.68 30.44 29,65
day 28.57 28.01 29.69 28.95 26.91 30.05 31.17 30.81 30.86 29.47 29.00 31.14 29.55
Wakulla 29.54 28.65 2825 29.46 27.20 32.03 33.32 29.55 30.71 28.45 29.64 27.80 29.53
Citrus 31.37 28.69 29.64 29.12 25.39 29.41 30.02 30.50 31.02 29.91 29.08 30.28 29.54
Sumter 28.81 27.07 27.65 27.30 27.02 30.61 32.91 31.26 32.93 29.03 29.25 29.17 29A2
Bay 29.13 29.83 28.57 28.39 27.92 30.11 31.28 29.34 29.46 28.02 28.92 28.49 29.12
Martin 27.90 27.32 2938 28.06 27.64 29.61 31.38 29.87 29.91 30.20 28.37 29.45 29.09
Nassau 29.87 29.02 3037 25.19 26.05 29.27 31.14 28.36 30.35 30.46 27.77 31.02 29.07
Indan River 27.91 27.65 26.93 29.62 28.03 29.61 31.65 30.14 29.85 27.08 30.26 28.22 2891
Gilchrist 31.07 25.92 29.64 28.64 25.26 30.85 30.30 29.07 28.82 27.91 29.29 27.39 28.68
Washington 30.20 26.88 26.77 29.20 30.30 29.27 29.73 27.46 31.29 25.77 28.96 28.13 28.66
Columbia 27.91 27.80 29.41 29.52 27.31 30.03 31.13 28.25 29.00 27.94 27.20 28.11 28.63
Levy 28.46 28.66 2846 28.72 25.68 30.00 30.25 28.70 29.28 27.58 28.75 28.28 28.57
Putnam 28.74 28.99 30.50 27.29 27.94 26.49 28.22 27.92 30.24 27.34 28.70 28.48 28.40
Monroe 27.00 28.17 30.09 27.00 26.84 28.24 29.50 29.82 29.13 25.76 29.18 27.75 28.21
Highlands 29.87 29.04 28.05 28.32 25.66 27.90 29.75 27.68 29.79 27.04 27.39 27.78 28.19
Hendry 30.64 28.67 26.48 28.40 26.54 27.40 30.46 27.45 29.28 27.48 25.36 27.45 27.97
Franklin 29.40 27.54 28.84 28.69 28.49 24.46 29.79 26.02 28.76 27.15 28.86 26.93 2791
Suwannee 26.76 27.52 29.16 28.28 25.99 28.87 27.76 28.96 28.74 27.31 27.91 27.48 2790
Dixie 24.64 29.54 30.05 28.42 24.06 27.56 28.20 26.94 28.46 28.98 29.05 28.68 27.88
Baker 27.72 27.48 26.15 27.83 25.47 27.88 30.17 27.96 28.84 28.56 27.83 28.39 27.86
Lafayette 27.00 26.80 25.39 26.85 27.64 28.51 28.43 27.77 31.19 27.00 27.36 28.35 27.71
Hardee 29.39 27.32 28.72 28.06 25.73 27.51 26.61 27.61 28.88 26.28 26.24 29.98 27.69
Jackson 28.57 27.53 28.04 26.58 25.28 27.68 29.08 27.90 27.93 27.52 27.39 28.15 27.64
Gadsden 27.82 27.48 26.61 26.98 25.30 28.49 28.64 28.94 29.58 26.05 26.88 27.83 27.55
Holmes 26.93 26.99 27.57 27.66 25.97 27.72 27.94 27.61 29.87 27.19 26.02 28.43 27.49
liberty 29.91 27.53 2598 30.81 25.79 26.96 30.39 22.48 2933 25.45 29.90 24.93 27.45
Calhoun 27.64 25.83 28.11 29.18 25.43 29.18 30.17 24.18 28.93 26.00 25.43 29.27 27.44
Jefferson 27.65 28.74 26.89 26.07 2601 2730 26.88 27.95 28.79 26.10 28.04 28.14 2738
Desoto 29.83 25.63 26.46 28.63 25.13 27.56 29.89 26.52 28.13 26.54 26.00 27.89 2735
Glades 27.28 27.97 24.02 27.10 24.69 30.94 28.63 27.60 28.81 26.21 27.31 27.14 2731
Okeechobee 27.13 27.10 27.46 26.47 25.13 29.09 28.55 25.80 28.98 26.72 26.32 28.03 27.23
Madison 27.14 26.85 26.79 26.61 26.23 28.21 28.98 25.29 29.63 28.15 25.39 27.50 27.23
Bradford 26.41 27.61 26.69 28.20 26.61 27.46 29.06 26.28 27.23 25.86 27.46 26.70 27.13
Union 29.06 24.85 28.40 27.63 24.26 25.73 30.74 25.80 29.11 25.38 27.28 26.39 27.05
Taylor 24.25 28.00 23.46 27.40 29.46 26.67 27.87 25.52 27.99 27.37 25.92 27.78 26.81
Gulf 26.62 27.61 28.67 25.26 25.92 2635 27.41 23.37 24.63 28.22 30.46 25.52 26.67
Hamilton 26.64 26.88 21.76 26.02 27.87 29.47 27.03 25.58 28.27 25.32 24.14 27.59 2638
Average 34.06 33.46 34.11 33.48 31.82 35.20 37.32 35.94 36.77 35.48 35.05 35.06 34.81

aCounties are sorted by total average momentum.








61




Table 3-5. Momentum index by county for years 1991 to 2002.


1993 1994 1995 1996 1997 1998


County,
Dade
Broward Orange Palm Beach Hillsborough Duval Pinellas Lee
Polk Brevard Seminole Collier Volusia Sarasota Mario Manatee Osceola Pasco
Lake Escambia Saint Johns Leon Waltco Alachua Okaloosa Santa Rosa Hernando Flagler Charlotte Saint Lucie Clay
Wakulla Citrus Sumter Bay
Martin Nassau Indian River Gilchrist Washington Columbia Levy Putnarn Monroe Highlands Hendry Franklin Suwannee Dixie Baker Lafayette Hardee Jackson Gadsden Holmes Liberty Calhoun Jefferson Desoto Glades Okeedobee Madison Bradford Union Tayl or Gulf
Hamilton


1991 105.28
104.54 104.18 104.08 103.98 103.79 103.76
103.40 103.28
103.41 103.15
103.24 103.30 103.11 103.03 103.18 102.95
103.24 102.94 103.00
102.94 103.09 103.03 102.97 103.11 103.13 102.85 102.92
103.12
103.14 102.86 102.95
103.14 102.88 102.91 102.79 102.99 102.79 103.11 103.02 102.79 102.85 102.87 102.70 102.99 103.06
102.94 102.68
102.46 102.77
102.70
102.94 102.86 102.78 102.69 102.99 102.76 102.77 102.98 102.73 102.71 102.71
102.64 102.91
102.42 102.66 102.66


1992 110.36
109.13
108.22 108.11 10798 107.52
107.41 106.56
106.44 106-76 106.30 106.28
106.40 106.16 106.13 106.17 106.07 106.17 10595
106.08 105.87 106.09 10599 10598
106.06 106.28 105.85 105.71 10595 10596 105.66
105.82 106.01 105.59 10590
105.52 105.89 105.56
105.70 105.71 105.57
105.71 105.77 105.52
105.89 10593
105.69
105.43 105.42 105.52 10538
105.67 105.61
105.53 105.39
105.74 10535
105.64 105.55
10552
105.42 105.40
105.40 10539
105.22
105.42 10535


'Counties are sorted by 2002 momentum index.


116.07 121.27 113.86 118.57 112.37 116.17 112.32 116.37 111.93 115.98 111.48 115.19 111.14 114.81 110.04 113.47 109.62 112.88 109.99 113.16 109.50 112.50 109.61 112.71 109.59 112.68 109.24 112.39 109.21 112.36 109.18 112.33
109.14 112.10 109.12 112.06 108.98 111.96 109.13 112.17 108.84 111.81 109.10 112.13 108.84 111.68 108.96 111.90 109.16 112.06 109.22 112.19 108.77 111.91 108.80 111.84 108.86 111.72 108.81 111.77 108.63 111.52 108.64 111.59 108.97 111.88 108.35 111.08 108.75 111.59 108.46 111.27 108.93 111.45 108.25 111.21 108.66 111.53 108.38 111.31 108.51 111.46 108.56 111.43 108.82 111.55 108.52 111.22 108.70 111.53 108.58 111.42 108.58 111.45
108.34 111.17 108.42 111.26 108.14 110.92 107.94 110.62 108.54 111.35 108.41 111.07 108.19 110.89 108.15 110.92
108.34 111.42 108.16 111.08 108.33 110.94 108.19 111.06 107.93 110.64 108.17 110.82 108.08 110.74 108.07 110.89 108.23 110.99 107.57 110.31 108.29 110.82 107.53 110.13


12632 12291 120.17
120.28 119.80
118.94 11830 116.70 115.83 116.08
115.44 115.68 115.67 115.26 115.21 115.25
11490
114.81 114.80 11494
114.70 11498 114.50
114.69 115.00
11493
114.64 114.74 114.40 114.55 114.21 114.31 114.42 113.79
11438 114.03 114.05 114.01 114.05
11434 114.20 114.00 11435 113.91
114.09 114.07 114.30 113.77 113.67
113.47 113.39 11392 113.60
113.42 113.51
114.00 113.62
113.54 113.57
113.11 11333 113.36 113.55
113.42 113.26 113.41
11292


131.77 137.51 127.69 132.71
124.69 129.45 124.66 129.19 124.15 128.68 122.99 127.45 122.16 126.28 120.14 123.86 119.20 122.83 119.43 122.95 118.96 122.58 118.96 122.53 118.93 122.38 118.50 122.01 118.38 121.86 118.32 121.61 118.03 121.22 118.00 121.34 117.91 121.27 118.13 121.44 117.65 121.11 118.09 121.26 117.59 120.82 117.76 120.90 118.06 121.08 118.04 121.25 117.59 120.95 117.75 120.88 117.37 120.51 117.48 120.52 117.22 120.33 117.51 120.84 117.36 120.36 116.85 120.14 117.39 120.52 116.99 120.13 116.98 120.09 116.98 120.14 117.14 120.17 117.26 120.24 117.20 120.31 117.00 120.02 116.99 119.82 116.73 119.68 116.88 119.86 116.81 119.86 116.74 119.72 116.66 119.43 116.43 119.25 116.25 119.27 116.24 119.08 116.67 119.33 116.37 119.27 116.27 119.13 116.28 119.08 116.70 119.74 116.54 119.55 116.27 118.96 116.32 119.31 116.20 119.06 116.24 119.09 116.18 119.08 116.30 119.20 115.99 119.07 115.92 118.71 116.04 118.78 115.86 118.57


143.27 137.62
134.07 133.74 133.27 131.55 130.23
127.44 126.29 126.20
125.94 126.04 125.58 125.31 125.22
124.64 124.62 124.42 124.45 124.54 124.29
124.38 123.95 123.99
124.28 124.29 123.95 123.85 123.55
123.49 123.41 123.80
123.41 123.26
123.46
123.12 122.93 123.15 123.07
122.98
123.14 122.89 122.61 122.67 122.63 122.61 122.32 122.33
121.94 122.07 121.86 122.09 122.06 122.03
121.84 121.98 121.97 121.75 121.97
121.82 121.67 121.61 121.83 121.65 121.26
121.12 121.13


1999
149.02 142.72 138.78 138.23 137.95 135.67
134.18 131.17 129.82 129.68 129.58 129.52 128.99 128.67 128.66 127.97
128.02 127.76 127.82 127.77 127.53 127.62 127.16
127.14 127.32 127.18 126.90 127.03 126.65 126.56 126.50 126.87 126.52 126.56
126.40 126.11 125.96
126.14 125.96 126.11
126.04 125.82 125.63 125.58 125.61 125.53 125.20 125.20
124.79 124.95 124.98 124.98 124.86 124.98 124.83 124.92 124.86 124.63 124.78 124.70 124.57 124.57 124.56 124.56 124.06 123.58 123.95


2000 2001
154.78 160.46 147.84 152.92 143.44 147.91 142.89 147.33 142.28 146.67 139.68 143.70 138.23 142.13 134.76 138.36 133.39 136.61 133.24 136.58 132.88 136.14 132.89 136.27 132.36 135.58 131.93 135.25 131.77 134.83 131.12 134.23 131.07 134.08 130.82 133.98 130.88 133.95 130.81 133.78 130.63 133.62 130.56 133.53 130.11 133.15 130.08 133.17 130.26 133.20 130.12 133.10 129.81 132.74 129.96 132.85 129.61 132.68 129.57 132.53 129.45 132.35 129.72 132.68 129.51 132.41 129.46 132.39 129.21 132.10 129.13 131.96 129.01 131.78
128.85 131.87 128.75 131.68 128.69 131.58 128.83 131.55 128.58 131.45 128.37 131.24 128.15 131.07 128.31 131.05 128.28 130.82 127.91 130.80 127.94 130.73 127.68 130.59 127.81 130.59 127.68 130.41 127.61 130.23 127.61 130.35 127.59 130.28 127.55 130.15
127.46 130.45 127.46 130.01 127.24 130.04 127.43 130.03 127.32 130.05 127.24 129.87 127.39 129.93 127.14 129.89 127.10 129.82 126.80 129.39
126.41 129.45 126.48 128.90


2002
165.80 157.71
152.24 151.70 151.11
147.64 145.92 142.03 140.03 139.82 139.60
139.43 138.80
138.48 138.01 137.43 137.32 137.23 137.03 136.88 136.81 136.60
136.28 136.15 136.12 136.03 135.76 135.73 135.69 135.58 135.46 135.46
135.44 135.30
134.95 134.91 134.89 134.70 134.42 134.40 134.36 134.28 134.08 133.85 133.83 133.56
133.49 133.47
133.46 133.43 133.25
133.23 133.16 133.06 132.99
132.94 132.93 132.86 132.82 132.77 132.68 132.68 132.56
132.46 132.17 132.00 131.66


121.13









62


170 160 150






*140


E










120 110 100
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Year


Figure 3-4. Momentum index of each Florida County plotted for years 1991 to 2002.













CHAPTER 4
METHODOLOGY OF COMPARATIVE VALIDATION

This chapter discusses all of the steps taken during our research to comparatively validate the six different forecasting techniques identified in Chapter 2. Chapter 3 outlined how momentum theory was derived and applied to strategic construction-market forecasting. This chapter applies the five remaining alternative research methods. These techniques include; 1), direct forecasting, 2) factor analysis, 3) multivariate-regression, 4) MERIC economic momentum analysis, and 5) gap analysis. A one, two and three year trend projection analysis is completed for the forecasting methods. Next, the forecasted results from the alternative methods and the momentum analysis are comparatively validated against a county's actual construction activity using simple regression. Next, all 67 Florida counties are rank ordered using their output variables and compared to their actual nonresidential permit (NRPERMIT) rank order. Finally, cluster analysis is used as a way to group the counties and is compared to two other classification techniques.

There are four general steps to the comparative validation outlined in this chapter. These four steps are summarized in Figure 4-1.


1 2 3 4
Compare Compare
Apply five forecasts variance of Cluster
forecasting with actual forecasted analysis
methods construction L and actual comparision
activity rank order

Figure 4-1. Four steps for comparative validation of research methods.


63






64


Alternative Research Approaches

As discussed above, five alternate methods were used in our research to forecast a county's nonresidential construction activity. The following sections of this chapter describe the application and methodology of each of these approaches. Direct Forecasting

The first alternative forecasting approach used to predict a county's nonresidential construction activity is directforecasting. As discussed in Chapter 2, direct forecasting is the simplest of the forecasting methods and is the only single variable qualitative method presented in our research.

Applying the technique of direct forecasting to strategic construction-market

forecasting is quick and simple. The direct forecast, or estimate, is normally made from a conscious or unconscious evaluation of a very small number of key demand variables that are known to most influence the required estimate. In the case of construction-market forecasting, these estimates are based on the key indicators that would most highly correlate to a county's construction activity.

There are six variables that were chosen for use with the direct forecast

methodology. These six forecast variables are listed in Table 4-1. The first five variables selected are the variables from each variable group outlined in Chapter 2 that best correlated with the dependent variable nonresidential permit (NRPERMIT). These variable correlations were previously calculated during the momentum analysis in Chapter 3.

The sixth variable chosen for use is the total annual value of residential permits

issued in a county (RPERMIT). RPERMIT data has historically been collected and used by the industry as one of the best indicators of overall construction activity for local,






65


state, and national construction-markets. The independent variable RPERMIT will be used in our research only as a variable in this directforecast methodology. The RPERMIT variable has be excluded from the momentum analysis and all other methods in our research.

Table 4-1. Six variables used in direct forecasting methodology.
No. Variable Code Variable Name
1 POP Total Population
2 DVMT Daily Vehicle Miles Traveled
3 CPAYROLL Construction Payroll
4 TOTEMPLY Total Employment
5 TOTTAX Total Taxes
6 RPERMIT Residential Permits

Methodology. The actual values from the six variables selected above were used as the direct forecasts, hence, a direct forecast. A rolling one, two and three year trend projection analysis was completed for each of the six variables in all of the Florida counties. This trend projection analysis methodology is described later in this chapter. The forecasted values of the six variables were compared against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6. Factor Analysis

The second alternative forecasting method used to predict a counties construction activity isfactor analysis. As discussed in Chapter 2, factor analysis is a generic name given to a class of multivariate statistical methods whose primary purpose is to define the underlying structure in a data matrix.' The variable data is condensed into a smaller set of factors. These factors can then be substituted for the original variables. These factors ' Hair, J. F. et al. (1998). (p. 90).






66


can then be objectively compared against the results of the other forecasting methods. The statistical software SPSS Base version 11.5 (2002) was used to perform the factor analysis on the datasets.

Methodology. The factor analysis was completed with the same ten independent variables used in the momentum analysis outlined in Chapter 2. The first step was to calculate the factor loadings for all counties. These factor loadings are linear combinations of the variables for each county and serve as an individual description of the variables. The factor loadings were then Varimax rotated and normally converged within 2 to 3 rotations. The Latent Root Criterion with an Eigen value of greater than 1 was used to determine the number of factors to extract. The set of factors explaining the greatest amount of variance in the variables was used.

The computed factor scores for each county were saved as the independent

variables to forecast a county's construction activity. A rolling one, two and three year trend projection analysis was completed for the computed factor scores. This trend projection analysis methodology is described later in this chapter. The forecasted values of the factor analysis were compared against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6. Multivariate-Regression

The third alternative forecasting method used to predict a counties construction

activity is multivariate-regression. Regression analysis is probably the most widely used and versatile dependent forecasting technique. Multivariate-regression analysis is a general statistical technique used to analyze the relationship between a single dependent






67


(criterion) variable and several independent (predictor) variables.2 The objective of this method is to use known independent variables to predict a desired dependent variable. A weight for each independent variable is calculated by the regression analysis to ensure the best prediction from the set of independent variables. The weights denote the relative contribution of the independent variables to the overall prediction and facilitate interpretation as to the influence of each variable in making the prediction.3 The statistical software SPSS was used to perform the multivariate-regression on the datasets.

Methodology. The multiple regression equation used for our research is shown in Equation 4-1.

(Eq. 4-1) Y = 00 + 01XI + 02X2 .+ OXIO

The actual nonresidential permit values (NRPERMIT) were used as the dependent variable (Y). The same ten independent variables used in the momentum analysis were used for the regression independent variables (X1..10). The model coefficient f0 is the intercept and OI..n is the model slope. The Enter regression method was used with a probability ofF entry .05 and removal of .10. The regression analysis was completed for all 67 counties during the years 1990 through 2002, and the model coefficients (O's) were calculated.

The resulting model was used to predict the value of NRPERMIT for all of

Florida's counties during the same time period. A rolling one, two and three year trend projection analysis was completed for the forecasted values. This trend projection analysis methodology is described later in this chapter. The forecasted values of the


2 Hair, J. F. (p. 142). 3 Hair, J. F. (p. 148).






68


multivariate-regression were compared against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6. MERIC Economic Momentum Analysis

The fourth alternative forecasting method used to predict a counties construction

activity is MERIC economic momentum analysis. As discussed in Chapter 2, the MERIC methodology measures economic momentum in a county relative to the overall economic momentum of a state. This index is a composite of percentage changes in personal income (TOTPINC), population (POP), and employment (TOTEMPLY) at the county level. An index equal to 0 means the county realized average economic growth during the decade. An index less than zero indicate relatively sluggish growth, while an index greater than zero indicates relatively prosperous growth.

Methodology. The first step of the MERIC method is to calculate the annual percentage change in the TOTPINC, POP and TOTEMPLY variables for each county using example Equation 4-2.

(Eq. 4-2) Annual percent change = ((Year 2001 - Year 2000) / Year 2000)

This percentage change was then standardized across all counties using Equation 4-3.

(Eq. 4-3) Z =


In Equation 4-3, X is the value you want to normalize, "A" is the arithmetic mean of the distribution, and "a" is the standard deviation of the distribution.

Finally, the three standardized variable estimates (i.e., TOTPINC, POP and

TOTEMPLY) were then averaged for the final index score. A rolling one, two and three year trend projection analysis was completed for the forecasted scores. This trend






69


projection analysis methodology is described later in this chapter. The forecasted values of the MERIC methodology were compared against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6. Gap Analysis

The final alternative forecasting approach used to predict a counties construction activity is an adaptation of the Expenditure-Sales Gap Analysis that was discussed in Chapter 2 of our research. The essence of Gap Analysis is to find some way of comparing supply and demand.4 This comparison was achieved by comparing a county's construction resources (supply) with its construction activity (demand). The level of construction resources was measured by a county's total Construction Payroll (CPAYROLL) and the level of construction activity by a county's total construction permits (NRPERMIT).

Methodology. The value of NRPERMIT generated for each dollar of CPAYROLL was calculated statewide and for each county by dividing NRPERMIT by CPAYROLL. Next, a county's expected payroll was calculated by dividing the NRPERMIT for each county by the average statewide NRPERMIT generated per each dollar of CPAYROLL. Finally, the gap was calculated by subtracting the actual payroll from the expected payroll. An example calculation of this gap analysis is shown in Table 4-2.








4 Clapp, J. M. (1987). (pp. 182).






70


Table 4-2. Example gap analysis calculation
Gap Analysis Calculation USD$
A Total value of construction permits (NRPERMIT) - Statewide $20,000,000
B Total construction employment payroll (CPAYROLL) - Statewide $1,000,000
C NRPERMIT generated for each dollar of CPAYROLL (C = A / B) $20

D Total value of construction permits (NRPERMIT) - County 1 $1,200,000
E Total construction employment payroll (CPAYROLL) - County 1 $40,000
F NRPERMIT generated for each dollar of CPAYROLL (F = D / E) $30

G Expected CPAYROLL for County 1 (G = D / C) $60,000
H Actual CPAYROLL for County 1 (H = E) $40,000
I GAP (additional construction payroll needed to meet construction activity demand) $20,000

The counties were then classified by their positive gap, balanced gap, or negative gap. As discussed in Chapter 2, a positive gap indicates that additional construction resources are needed in the county to meet the construction activity demand. Balanced gap indicates a balance between the supply of construction resources and the demand of construction activity. A negative gap indicates a surplus of construction resources exist for the corresponding demand for construction activity. The three classifications of gap and their corresponding definitions are shown in Table 4-3. The findings of this gap analysis calculation and classification are included in Chapter 5. Table 4-3. Payroll gap type and definitions.
Gap Definition
(Gap as a percent of actual Gap
Gap Type Description statewide CPAYROLL) Code
Positive Demand > Supply Gap > 0.25% 2
Balanced Demand = Supply 0.25% Gap -0.25% 1
Negative Demand < Supply Gap < -0.25% 3

In the final step of the gap analysis, the counties were sorted in descending order based on the absolute value of the gap. This gap value was used to predict the level of construction activity in a county. A rolling one, two and three year trend projection analysis was completed for the absolute value of the gap. This trend projection analysis methodology is described later in this chapter. The forecasted values of the gap analysis






71


were compared against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6.

Comparative Validation of the Forecasting Methods

Thus far, a total of six forecasting methodologies have been used in our research to measure and forecast a county's nonresidential construction activity. Chapter 3 detailed a new methodology of momentum analysis (including opportunity index and slope). This chapter has detailed five alternative forecasting methods including; 1), direct forecasting, 2) factor analysis, 3) multivariate-regression, 4) MERIC economic momentum analysis, and 5) gap analysis. These six methods have provided a total of twelve individual forecasts. These variables and their associated forecasts are shown in Table 4-4. The question now becomes which of these forecasting methods best predicts a county's future nonresidential construction activity (i.e., NRPERMIT). The remaining sections of this chapter will compare and statistically validate all six of these forecasting methods. Trend Analysis of Key Construction Indicators

As mentioned in previous sections of our research, a rolling one, two and three year trend projection analysis was completed for the twelve output variables from each of the six forecasting methods. The purpose of these trend projections is to test the various methodologies for any inherent advantages relating to the duration of the forecast. The linear trend of each output variable in a specific year was measured and projected using the method of least squares. This methodology fits a straight line to the variables over time. A minimum of five (5) known data points were used to forecast values one year, two years and three years out. The forecasted values of all methods were compared






72


against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the variance were analyzed and the findings are included in Chapter 6. Table 4-4. Six forecast methodologies and their twelve associated output variables. Method Output Variable Variable Description
Momentum analysis OINDEX Momentum index
OSLOPE Momentum index slope
Direct forecasting POP Population
TOTTAX Total tax
CPAYROLL Construction payroll
TOTEMPLY Total employment
DVMT Daily vehicle miles traveled
RPERMIT Residential permits
Factor analysis FACTOR Factor regression output variable
Multivariate-regression LINEAR Multivariate-regression output variable
MERIC economic MERIC MERIC forecast output variable
momentum analysis
Gap analysis GAP Gap forecast output variable

Methodology for Comparative Validation

The purpose of the following statistical analysis is to compare a county's actual construction activity to a county's forecasted construction activity and validate the accuracy of each method. Simple regression was used for this purpose.

Simple regression is similar to the multivariate-regression method discussed earlier in this chapter. The difference is that the simple regression methodology involves the analysis of a single dependent variable (NRPERMIT) and its relationship to a single independent metric variable (forecasted activity). Simply stated, the twelve individual forecasts from the six different methodologies were each tested individually against a county's nonresidential construction activity. The simple regression equation used for our research is shown in Equation 4-4. SPSS was used to perform the simple regression analysis on the data sets.

(Eq. 4-4) Y = fo + 1XI






73


The actual nonresidential permit values (NRPERMIT) were used as the dependent variable (Y). Each output variable calculated from the twelve previous forecasting methods was used for the independent variable (XI). These variables were presented earlier in Table 4-4. The Enter regression method was used with a probability of F entry .05 and removal of .10. The model coefficient flo is the intercept and fl is the model slope.

The regression analysis was completed for all 67 counties during the years 1996 through 2002. This methodology was used to select the forecasting method that had the highest correlation with the actual construction activity for the one year, two year, and three year trend projections. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6.

During this regression analysis, the MERIC and GAP methodologies consistently demonstrated a weaker correlation to the dependent variable NRPERMIT and were not included in the following county rank variance analysis or the cluster analysis. The results of these two methodologies are included in Chapter 6 County Rank Variance

The previous forecasting methods have allowed the rank ordering of the 67 Florida counties based on their output variables. These rank orders were compared to the counties actual NRPERMIT rank order. This comparison was completed for each year of the one, two, and three year trend projections and the variance was analyzed. This variance calculation measures the variance of a dataset population based on the entire population. During the preceding simple regression analysis, the MERIC and GAP methodologies consistently demonstrated a weaker correlation to the dependent variable NRPERMIT and were removed from this variance analysis. The statistical variance of






74


the ten remaining method output variables was analyzed and the findings are included in Chapter 6.

Cluster Analysis

Overview

Cluster analysis was previously selected in Chapter 2 as one of the interdependent regression techniques to be used in our research. Cluster analysis is a multivariate procedure whose purpose is to group objects based on the characteristics they possess. Cluster analysis classifies objects (i.e., Florida counties) so that each object is very similar to others in the cluster with respect to the independent variables. The resulting clusters of objects (counties) should then exhibit high internal (within-cluster) homogeneity and high external (between-cluster) heterogeneity.5 The intent of cluster analysis is the comparison of objects based on the independent variables, not an estimate of the variate. Further, the researcher is searching for the natural structure among the observations based on a multivariate profile.6

There are two primary methods of cluster analysis: (a) Hierarchical Cluster

Analysis and (b) K-Means Cluster Analysis.7 The k-means method is intended to handle large research data sets with 200 or more cases, where the hierarchical method is designed for smaller data sets. Because our research includes only 67 counties, the hierarchical method was chosen.




5 Hair, J. F. (p. 473).

6Hair, J. F. (p. 470).

7Statistical Package for the Social Sciences, Inc., Base 10.0 Applications Guide. (1999). Chicago, IL: SPSS Inc. (p. 293).






75


Cluster analysis was used for several purposes in our research. The primary purpose was to cluster, or group, the counties by their variable characteristics and to compare these clusters to other classification techniques. Next, cluster analysis was used to evaluate the relationships between the county clusters and each individual variable (key indicator). The goal was to see if certain key indicators are better at predicting different clusters of counties. Finally, cluster analysis was used to evaluate the relationships between the county clusters and each forecasting method. The goal of this analysis was to see if certain forecasting methodologies are better at predicting different clusters of counties.

Methodology for Cluster Analysis

Our research has generated two primary data sets. The first variable data set

includes the sixteen original variables for the 67 Florida counties over a thirteen year time period. The second methodology data set includes the twelve forecasted output variables from the six different methodologies. The variable data set was used in the cluster analysis to group the counties.

Due to the overall consistency in the annual trends of the research variables, and the scope limits of our research, the cluster analysis was only completed on the latest 2002 variable data set. SPSS was used to perform the cluster analysis on the variable data set. The cluster analysis was completed with the same ten independent variables used in the momentum analysis. The cases were clustered and labeled by county. A range of 2 to 8 solutions for the cluster membership was requested. Both a Dendrogram and an icicle plot were requested up to the 8th cluster range. A between-groups linkage cluster method was used with the distance measured by the Pearson's correlation interval. The measures were transformed to absolute values. The agglomeration schedule, cluster






76


membership table, Dendrogram and the icicle plots were analyzed to identify the most appropriate county clusters.

Construction-market classification

The cluster analysis output showed that the counties could be classified into three distinct groups, or cluster memberships. Further review of the cluster analysis output and plots showed that cluster 3, the largest cluster of counties, could be broken down into three sub-clusters. This provided a total of 5 county clusters for comparison purposes. A county's cluster membership (i.e., I through 5) was compared to two other classification methods. These methods included gap classification (1, 2, or 3) previously outlined in this chapter, and market share classification which is outlined in the following paragraphs.

A county's share of the total nonresidential construction-market was compared to a county's cluster membership and gap classification. A county's market share was calculated by dividing the total nonresidential construction activity in a specific county by the total nonresidential construction activity in the State of Florida. The counties were then classified into three market share groups. These three groups and their corresponding definitions are shown in Table 4-5. Table 4-5. County market share classifications.
Market Share Definition
Market Share Level (Percent of statewide NRPERMIT) Code
High > 5.0% 3
Medium 1.0% to 5.0% 2
Low < 1.0% 1

The results of the gap, cluster and market share analysis were statistically compared using the Pearson correlation coefficient identified in Chapter 3. A two-tailed test of significance was used. The findings of this comparison are discussed in Chapter 5.





77


Key indicators and cluster comparison

Next, cluster analysis was used to evaluate the relationships between the county clusters identified above and each key indicator identified in Chapter 3. The Pearson correlation coefficient was computed between thefive county clusters and the same ten key indicators used in the momentum analysis. Further, nonresidential and residential construction activity was also analyzed. A two-tailed test of significance was used. The findings of the key indicator and cluster comparison are discussed in Chapter 5. Forecasting methods and cluster comparison

Each output variable from the different forecasting methodologies was also tested against the five county clusters. The simple regression equation used for this test was shown previously in Equation 4-4. SPSS was used to perform the simple regression analysis on the data set.

The actual nonresidential permit values (NRPERMIT) were used as the dependent variable. As discussed earlier in this chapter, the MERIC and GAP methodologies consistently demonstrated a weaker correlation to the dependent variable NRPERMIT and were not included in this analysis. The remaining ten output variables calculated from the forecasting methods were used for the independent variable. These variables were presented earlier in Table 4-4. The Enter regression method was used with a probability of F entry .05 and removal of .10. The regression analysis was completed for all five clusters of counties for the year 2002 only. The findings of the forecasting method and cluster comparison are discussed in Chapter 6.














CHAPTER 5
KEY INDICATOR FINDINGS

This chapter presents the research finding related to the key indicators of

construction. The chapter begins with a review of the overall results for the key indicator analysis. This is followed by a discussion of the findings for each key indicator construct and its associated variables. Next, the results from the three construction-market classification methodologies are reviewed and classification maps are presented. Finally, the findings regarding the relationships between the key construction indicators and the county clusters are presented. In the next chapter of our research, the findings related to the various forecasting methods are presented.

Overview of Key Indicators Results

A total of sixteen key construction indicators were initially selected for our research in Chapter 2. The key construction indicators were grouped into six independent variable constructs. These independent variables and their associated constructs are shown in Table 5-1.

Table 5-1. Key construction indicator constructs and associated variables.
Population Geographic Initial Employment Economic Financial
Advantage Infrastructure Transition Environment Resources
POP PROX DVMT CPAYROLL TOTEMPLY TOTTAX
PDENSITY CLMILES TOTPINC GSALES ALVALUE
RDENSITY AVEWAGE PLINDEX TOTREV
RPERMIT

In Chapter 3, the Pearson correlation was computed for all of the independent

variables for the years 1990 through 2002. The results of this analysis are summarized in Table 5-2. The indicators have been sorted by their average correlation to NRPERMIT


78





79


and include the variable description and group. Eleven of the 16 variables were highly

correlated (> 0.800) to the dependent variable nonresidential permit (NRPERMIT).

These eleven variables have Pearson's Coefficient averages ranging from 0.829 to 0.962.

The remaining five variables were then dropped from use in the research due to their low

correlation (< 0.800) to NRPERMIT. These five excluded variables are shown below the

dashed line in Table 5-2.

In addition to these five excluded variables, the independent variable RPERMIT

was used in our research only as a variable in the directforecast methodology. The

RPERMIT variable was excluded from the momentum analysis and all other methods in

our research.

Table 5-2. Pearson correlation results for each research variable.
Variable Average Variable Group
Correlation
NRPERMIT 1.000
TOTTAX 0.962 Financial Resources
ALVALUE 0.960 Financial Resources
TOTEMPLY 0.948 Economic Environment
GSALES 0.941 Economic Environment
CPAYROLL 0.940 Employment Transition
DVMT 0.932 Initial Infrastructure
POP 0.924 Population
TOTPINC 0.912 Employment Transition
TOTREV 0.856 Financial Resources
RPERMIT 0.854 Initial Infrastructure
CLMILES_____ 0.829 Initial Infrastructure
AVEWAGE 0.660 Employment Transition
PLINDEX 0.637 Economic Environment
PDENSITY 0.624 Population
RDENSITY 0.458 Initial Infrastructure
PROX 0.360 Geographic Advantage

The following sections of this chapter analyze the key variable correlations to

NRPERMIT by using the six independent variable constructs previously discussed.

Remember that these constructs are nothing more than logical groupings of the key





80


construction activity indicators identified in the preceding literature review. As discussed in Chapter 2, using these constructs as descriptors of the relationship between the key indicators and a county's construction activity allows for a more understandable and vivid discussion.

Financial Resources

Overall, the financial resources group of key indicators was the most predictive of nonresidential construction activity. As shown in Table 5-2, two of the three financial resource variables, (TOTTAX and ALVALUE) had the highest correlation with NRPERMIT and the third finished in the top eleven. This is a fairly logical finding because the financial resources must be available to the marketplace to construct new facilities and infrastructure. As was suggested in Chapter 2, a market's access to financial resources is positively correlated with a county's construction activity. Economic Environment

The second most predictive group of key indicators appears to be related to a county's economic environment. As shown in Table 5-2, two of the five highest correlating variables were in this group. A county's economic environment must be conducive to the growth of investment and employment. The improvement of the economic environment within a county was found to be positively correlated with a county's construction activity.

One dissenting variable from this group was a county's price level index

(PLINDEX). The price level index is a set of numbers which reflects the price level in each county relative to population-weighted statewide average (100 for each category) for a particular point in time. It measures price level differences from place to place in contrast to the consumer price index prepared by the U.S. Bureau of Labor Statistics,





81


which measures price level changes from month to month.' The basket of goods measured in the price level index includes housing, food and beverage, health care, transportation and other miscellaneous goods and services. It appears that this variable is not applicable at the county level because it uniformly affects all of the counties in the state, or the effect of "a rising tide raises all boats." Employment Transition

The next most predictive group of key indicators appears to be employment transition. This group of variables is closely related to the economic environment construct discussed above and includes the county's transition into certain types of higher paying employment. As discussed in Chapter 2, shifts from the lower income employment sectors to the higher income sectors will stimulate investment in new facilities and infrastructure. This employment transition from low to high-income employment sectors was found to be positively correlated with a county's construction activity.

One dissenting variable from this group was a county's average wage

(AVEWAGE). Average wage is the average earnings per job in dollars. There is no clear reason for the lower correlation particularly when average wage was highly correlated with the best predictor, a county's total tax revenue (TOTTAX). Average wage seemed to predict best in the most populated counties. Initial Infrastructure

The next most predictive group of key construction indicators appears to be related to a county's initial infrastructure use and investment. Table 5-2 shows two of the three

1 Bureau of Economic and Business Research. (2002). Florida Statistical Abstract 2002. Gainesville: University of Florida, Warrington College of Business Administration. (p. 761).






82


variables correlating highly with NRPERMIT. As discussed in Chapter 2, public construction projects tend to become larger and more complex as a county develops. Population growth drives the demand for residential housing, water resources, and energy delivery systems. Unpaved roads are paved, and existing road capacity is increased. This initial construction activity is a predecessor to larger public construction projects such as toll highways, power plants, and water and wastewater treatment facilities. The increased use and growth of a county's roadway infrastructure was found to be positively correlated with a county's nonresidential construction activity.

One dissenting variable from this group was a county's road density (RDENSITY). Road density was computed by dividing a county's total centerline miles of roadway by its total land area (miles/square mile). The reason for this low correlation to NRPERMIT may be that a county's land area is simply the area within a geographic boundary and that this measure is independent and unrelated to either roadway miles or construction activity.

Population

Finally, and most unexpectedly, a county's total population (POP) was found to be only an average predictor of nonresidential construction activity in comparison to the other key variables. It was hypothesized that increases in a county's total population should drive the need for new buildings and infrastructure.

But why was the population variable correlation so low? One reason is that this variable may really be a function of the greater economic opportunity available in a geographic region, not the driver. A healthy economic environment generally creates more employment opportunity. This environment increases the demand for more and better skilled workers which in-turn drives higher wages. This opportunity and higher





83


income potential attracts population to the area. This was evidenced by the population (POP) variable's very high correlation to a county's total employment (0.98 1) and total personal income (0.979), but its lower correlation to NRPERMIT (0.924).

Further, a county's population density (PDENSITY) was not found to be a good predictor of construction activity for the same reasons previously discussed for road density.

Geographic Advantage

A county's proximity to major urban areas (PROX), or geographic advantage, was not found to be a good predictor of nonresidential construction activity. In Chapter 2, it was discussed how secondary urban areas, or bedroom communities, typically develop around larger cities. It was anticipated that this spillover effect would contribute to the development of adjacent counties and would be further amplified if the county is located between multiple large cities. To test this hypothesis, a proximity factor was constructed that measured the relationship of proximity and population between counties (see Chapter

2 for details). This factor provided a value that is weighted both by distance and population. While this factor did not predict well overall, it did perform well on the most populated counties, counties immediately adjacent to large cities, counties between larger counties, and counties adjacent to non-coastal urban centers like Orlando. There are several suggestions for improving this factor. First, using more than the top 10 most populated counties would allow counties such as Escambia (Pensacola) and Leon (Tallahassee) to be included in the calculation. This may help improve the results of the counties located throughout Florida's panhandle because the most populated counties are generally found in South Florida. Second, the calculation could be modified to include more than the 2 closest and most populated counties. This may help improve the results





84


of the counties that are not located directly between two large counties. Finally, the distance used to calculate the factor was the distance between the county seats. It may be more appropriate to use the distance between the centroids of a county's land area or population.

Construction-Market Classification

Three different methodologies were presented in Chapter 4 that can be used to

classify counties. These methods included cluster analysis, gap analysis and a county's percent share of the total nonresidential market. The classification results of these three methods are shown in Table 5-3 for all 67 Florida counties. To provide a geographical context for these findings, the results from Table 5-3 were mapped and are shown in Figure 5-1. The correlation matrix between these three methods is shown in Table 5-4.

The classification method inter-correlations ranged from 0.647 to 0.762. While these correlations may appear relatively low, it should be understood that these different classification methods are fundamentally different measures by the nature of their variables. Viewing the counties through these three classification methods has advantages over using just one classification method. The variables and advantages of each method are discussed in the following paragraphs.






85


Table 5-3. Results of the cluster, gap, and market share classification methods.


Cluster Analysis


Gap Type


County Membership Cod
Duval 5 3
Lee 5 3
Dade 5 2
Broward 4 3
Hillsborough 4 3
Orange 4 3
Collier 4 3
Pinellas 4 3
Seminole 4 3
Palm Beach 4 2
Manatee 4 2
Sarasota 4 1
Monroe 4 1
Brevard 3 3
Escambia 3 3
Lake 3 3
Alachua 3 2
Leon 3 2
Marion 3 2
Pasco 3 2
Saint Lucie 3 2
Polk 3 1
Volusia 3 1
Bay 3 1
Citrus 3 1
Clay 3 1
Desoto 3 1
Highlands 3 1
Okaloosa 3 1
Saint Johns 3 1
Osceola 2 2
Martin 2 1
Charlotte 2 1
Gulf 2 1


e


Percent Market Share
3
2 3
3 3 3
2 2 2 3
2 2 1

2 1
2 2 2 2 2 2 2



2 2 1

1


1


2 1 I


a Counties are sorted by cluster, then gap, then percent market share results.


County
Hardee Hendry Indian River Nassau Okeechobee Taylor
Flagler Santa Rosa Sumter Baker Bradford Calhoun Columbia Dixie Franklin Gadsden Gilchrist Glades Hamilton Hernando Holmes Jackson Jefferson Lafayette Levy Liberty Madison Putnam Suwannee Union Wakulla Walton Washington


Cluster Analysis Membership
2 2 2 2 2 2 1 1

I
1 1 1




1 1 1


1 1

1 1 1 1

1


Gap Type Code



2
1
1
1



2
2









1
1
1 1 1 1 1 1
1
1
1


Percent Market Share
1 I 1 I I 1 1 1 1 1 I 1 1 I 1 1 1 I 1 1 1 1 1 1 I 1 1 1 1 I 1 I 1















Table 5-4. Pearson correlation matrix for the three county classification methods.

Method Cluster Analysis Gap Analysis Percent Market Share

Cluster Analysis 1.000 0.647 0.762

Gap Analysis 0.647 1.000 0.731

Percent Market Share 0.762 0.731 1.000


Legend:
=Gp1Type: D= S - Gap Type 2: D > S Gap Type 3: D < S


(a) Cluster analysis


(b) Gap analysis


00


Legend:
- Market Sharc <1%
PMarken Shamk s% to 5% \
-M.rkct Share > 5%

-A
















(c) Percent market share.


Figure 5-1. Maps of county classification analysis. (a) Cluster analysis, (b) Gap analysis, and (c) Percent market share.


Legend:
S=Cluster I 0 -Cluster 2


S -Cluster 5




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EGGVHDLVJ_O6TUSP INGEST_TIME 2015-04-13T19:37:48Z PACKAGE AA00029869_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES



PAGE 1

COMPARATIVE EVALUATION OF STRATEGIC CONSTRUCTION-MARKET FORECASTING METHODOLOGIES By OTTO GEORGE FETTERHOFF III A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2004

PAGE 2

Copyright 2004 by Otto George Fetterhoffni

PAGE 3

This dissertation is dedicated to my lovely wife Michele and my two wonderflil sons Hans and Alexander. This endeavor would not have been possible without their unwavering patience, support, and love.

PAGE 4

ACKNOWLEDGMENTS First I would like to acknowledge William O'Brien for his support throughout the past four years. As my committee chair, he has provided the needed academic perspective and theoretical balance to what began as a rather pragmatic undertaking. Dr. O'Brien believes that identifying and applying an individual's inherent interests and abilities are a prerequisite to the success of any endeavor. I would like to acknowledge Marc Smith for serving as cochair of my committee. His fundamental knowledge of the subject matter and keen perceptiveness were the origin of significant portions of this work. I would like to acknowledge David Ling, Robert Stroh, and Charles Kibert for serving as members of my committee. Their open and insightful feedback was instrumental in assuring a quality and ecumenical outcome. Finally, I would like to acknowledge Brian Morris and the URS Corporation for making the temporal, financial, and other required resources available to complete this endeavor. Their support helped to minimize the burden of this undertaking on myself and my family. iv

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS iv LIST OF TABLES ix LIST OF FIGURES xii ABSTRACT xiii CHAPTER 1 INTRODUCTION 1 Issues Leading to this Research 2 Problem Statements 4 Objectives of the Research 4 Benefits and Significance of the Research 5 Organization of this Study 6 2 LITERATURE REVIEW 9 Management Approaches to Strategic Construction Marketing 9 Strategic Marketing Definitions 9 Strategic Marketing Growth Models and Planning Processes 10 Summary of Management Approaches 16 Key Indicators of Construction Activity 17 Construction-Market Segmentation 17 Trends and Forces in the Marketplace 19 Identification of Key Indicators 20 Environmental and Political Indicators 24 Key Indicator Selection and Constructs 26 Population 27 Geographic advantage 27 Initial infi-astructure 28 Employment transition 29 Economic environment 29 Financial resources 29 Summary of the Key Indicators 30 Approaches to Construction-Market Forecasting 31 V

PAGE 6

Qualitative Techniques 31 Quantitative Techniques 33 Regression analysis 35 Dependent regression techniques 36 Interdependent regression techniques 36 Trend analysis 37 Gap analysis 37 Law of Universal Gravitation 39 Momentum analysis 39 Momentum forecasting in the stock market 40 Economic momentum 41 Summary of Forecasting Approaches 42 Summary of Research Questions 43 3 MOMENTUM THEORY DERIVATION AND APPLICATION 45 Momentum Theory Introduction 45 Differentiation between Momentum and Regression Approaches 47 Data Sample and Data Sources 48 Momentum Theory Applied to Strategic Construction-Market Forecasting 49 Methodology for Analyzing County Momentum 49 Mass of key indicators 52 Velocity of key indicators 53 Influence of key indicators 54 Momentum of key indicators 55 Total momentum of a county 55 Derivation of the Momentum Index 56 Momentum Index Slope 58 Summary of Momentum Theory 59 4 METHODOLOGY OF COMPARATIVE VALIDATION 63 Alternative Research Approaches 64 Direct Forecasting 64 Factor Analysis 65 Multivariate-Regression 66 MERIC Economic Momentum Analysis 68 Gap Analysis 69 Comparative Validation of the Forecasting Methods '. 71 Trend Analysis of Key Construction Indicators 71 Methodology for Comparative Validation 72 County Rank Variance 73 Cluster Analysis 74 Overview 74 Methodology for Cluster Analysis 75 Construction-market classi fication 76 Key indicators and cluster comparison 77 vi

PAGE 7

Forecasting methods and cluster comparison 77 5 KEY INDICATOR FINDINGS 78 Overview of Key Indicators Results 78 Financial Resources 80 Economic Environment 80 Employment Transition 81 Initial Infrastructure 81 Population 82 Geographic Advantage 83 Construction-Market Classification 84 Key Indicators and County Clusters 88 6 FORECASTING METHODOLOGY FINDINGS 91 Statistical Overview 91 One, Two, and Three Year Trend Projections 92 Validation of Statistical Significance 94 Accuracy of Forecasting Methodologies 97 Direct Forecasting 97 Factor Analysis 98 Multivariate-Regression 100 Momentum Analysis 100 Gap Analysis 102 Missouri Economic Research and Information Center Economic Momentum Analysis 102 Where are the Best Construction-Markets? 103 County Rank Variance 104 Construction-Market Classification 108 7 CONCLUSIONS OF THE STUDY Ill Research Intent and Expectations 11 1 Momentum Theory 112 Comparative Validation of the Forecasting Methods 112 Research Results 113 Key Construction Indicators 113 Forecasting Methods 115 Methodology vs. complexity 115 Best construction-markets 117 Threats to Research Validity 118 External Validity 118 Construct validity 119 Statistical Validity 120 Internal Validity 120 Future Research 121 vii

PAGE 8

Key Indicators 121 Environmental and Political Influences 121 Momentum Forecasting Theory 122 Construction-Marketplace Life Cycle 122 Other Opportunities 124 APPENDIX A LIST OF POTENTL\L KEY CONSTRUCTION INDICATORS 125 B RESEARCH VARIABLE DATA FOR YEARS 1990 THROUGH 2002 146 LIST OF REFERENCES 160 BIOGRAPHICAL SKETCH 165 viii

PAGE 9

LIST OF TABLES Table page 2-1 Summary of construction activity variables listed in Appendix A 20 22 Key indicator constructs and associated variables 27 31 Research variable data, descriptions, and sources 51 3-2 Comparison of momentum theory variable definitions 52 3-3 Pearson Correlation Coefficient to nonresidential permit activity 57 3-4 Total momentum of a county for years 1991 through 2002 60 35 Momentum index by county for years 1991 to 2002 61 41 Six variables used in direct forecasting methodology 65 4-2 Example gap analysis calculation 70 4-3 Payroll gap type and definitions 70 4-4 Six forecast methodologies and their twelve associated output variables 72 45 County market share classifications 76 51 Key construction indicator constructs and associated variables 78 5-2 Pearson correlation results for each research variable 79 5-3 Results of the cluster, gap, and market share classification methods 85 5-4 Pearson correlation matrix for the three county classification methods 86 55 Results of the key indicator and cluster comparison for the year 2002 90 61 Forecast methodology comparison for 1, 2, and 3 year trend projecrions 93 6-2 Forecast methodology F & t statistic comparison for 1, 2, and 3 year trend projections 95 ix

PAGE 10

6-3 Forecast methodology rank and type comparison 99 6-4 Rank of nonresidential construction-markets as forecasted by total tax collections using a 3 year trend 106 6-5 Total county rank variance for forecast years 1996 through 2002 107 66 Results of forecasting methodology and cluster comparison 110 71 The relationship between the four construction-marketplace life cycle stages and marketing mix actions 123 A-1 Economic indicators 126 A-2 Construction and infrastructure indicators 128 A-3 Safety and health indicators 131 A-4 Education, social, and government indicators 133 A-5 Air environmental indicators 136 A-6 Water environmental indicators 138 A-7 Land environmental indicators 140 A-8 Ecology environmental indicators 141 A-9 Sound environmental indicators 142 A10 Natural resource environmental indicators 143 B1 1 990 variable data 1 47 B-2 1991 variable data 148 B-3 1 992 variable data 149 B-4 1993 variable data 150 B-5 1 994 variable data 1 5 1 B-6 1995 variable data 152 B-7 1 996 variable data 1 53 B-8 1 997 variable data 1 54 B-9 1998 variable data I55 X

PAGE 11

B-10 1999 variable data 156 B-1 1 2000 variable data 157 B-12 2001 variable data 158 B1 3 2002 variable data 159 xi

PAGE 12

LIST OF FIGURES Figure page 1-1 Organization of this study 3-1 Map of the 67 counties in the State of Florida 50 3-2 Seven general steps of momentum analysis 52 3-3 Three-dimensional relationships of momentum 56 34 Momentum index of each Florida County plotted for years 1991 to 2002 62 41 Four steps for comparative validation of research methods 63 51 Maps of county classification analysis 86 61 Map of the 2005 total tax-direct forecast results using a three year trend projection 105 6-2 County rank variance comparison for the year 2002 107 xii

PAGE 13

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy COMPARATIVE EVALUATION OF STRATEGIC CONSTRUCTION-MARKET FORECASTING METHODOLOGIES By Otto George Fetterhoff HI May 2004 Chair: William J. O'Brien Cochair: Marc T. Smith Major Department: Building Construction The relative historical stability of the U.S. economy and its strong influence on the construction industry have allowed large U.S. design and construction firms to naturally grow and adapt to the slow changes of the construction-market. But business changed suddenly for many large firms in 2001. The economic recession, combined with the market fallout fi-om the events of September 1 1, 2001, delayed or cancelled many design and construction projects. Despite these historic events, investor expectations prevailed, sending these firms searching to find new and alternative markets. This research sought to examine many of the challenges a marketing professional is confi-onted with when searching for new competitive markets. These challenges generally include questions regarding the key predictive indicators of construction activity, the methods of market forecasting, and the classification and selection of new potential markets. This research was also intended to initiate decisions regarding the spatial structure of a large design and constiiiction firm. xiii

PAGE 14

A review of the literature provided many different management approaches to strategic construction-market forecasting. Over 250 different indicators (in over 56 categories) were identified that could potentially influence construction activity within a given market. The literature review also provides an overview of the variety of forecasting methodologies that can be used to forecast and classify construction activity. Historical data were collected for all of Florida's 67 counties for the period 1990 through 2002. A new momentum forecasting methodology is presented that was derived fi-om Sir Isaac Newton's three laws of motion. This unique forecasting approach (along with five other existing approaches) is used to forecast construction-market activity at the county level. The best key indicators of future nonresidential construction activity at the county level were found to be the total annual tax collections, the total annual assessed value of commercial land, and the total annual number of wage and salary jobs in a county. The current industry practice of using housing starts as a key indicator was found to be one of the least accurate indicators of nonresidential construction activity in the State of Florida relative to the other methods tested. A direct forecasting methodology (using a single variable) was found to be the most accurate predictor of future nonresidential construction activity. This was the simplest of all the forecasting methodologies used, and consistently outperformed the more complex statistical multivariate-regression techniques. The new momentimi methodology was found to be a highly accurate alternative forecasting methodology that is less complex but more meaningful than the traditional statistical regression approaches. xiv

PAGE 15

CHAPTER 1 INTRODUCTION The general domain of the following research is market forecasting in the U.S. construction industry. More specifically, our research is intended for large design and construction firms working in the nonresidential construction-market. Our research applied several existing and one new forecasting methodology to the difficult management task of analyzing and prioritizing multiple construction-markets. Our research is intended to help these large firms focus their resources on the markets with the most opportunity. Geographical coverage is generally seen as a strategic business decision. Marketing and sales are organized mainly as a business process. Our research is intended to initiate decisions regarding the spatial structure of the organization. In short, the research goal is to locate the opportunity, not win an opportunity. A new forecasting method is presented that integrates Newton's natural science theory of momentum. This unique forecasting approach (along with several other existing approaches) is used to estimate nonresidential construction activity in all 67 counties of the State of Florida. A 13 year fime period was used for our study that began 1990 and ended in 2002. This momentum forecasting methodology is introduced as a more understandable way to conceptualize the complexities of strategic market forecasting. Finally, our research completed a comparative validation of the various forecasting techniques to confirm the accuracy of the different approaches. 1

PAGE 16

2 Issues Leading to this Research A cowboy heading westward came across an Indian lying in the middle of the wagon trail with his ear to the ground. The cowboy stopped and asked the Indian what he was listening to. Without getting up and his ear still to the ground, the Indian replied, "covered wagon with two calico horses heading west, large man with hat and rifle, woman in blue dress and two screaming children, wagon fully loaded with cast iron stove tied to back of wagon." Amazed at the details of the Indian's prediction, the cowboy asked the Indian how he could describe the wagon with such accuracy. The Indian replied, "Wagon ran over me thirty minutes ago." Author Unknown' Due to the relative historical stability of the U.S. economy and its strong influence on the construction industry, large design and construction (D&C) firms have naturally evolved, or organically grown' and adapted with the slow changes of the constructionmarket. But business changed suddenly for many large U.S. D&C firms in 2001. The economic recession combined with the market fallout fi-om the events of September 1 1, 2001, further delayed or in some cases cancelled the start of many construction projects in markets that were once thought to be secure. Despite the magnitude of these historic events, large D&C firms were soon pressured by their investors to return to the growth they knew before 2001 . These investor expectations forced many large D&C firms to evaluate alternative market areas that were historically not considered by their offices. The market area covered by a local D&C office may include a city, counties, a state, or multiple states. A market can be defined as a geographic or political boundary where D&C firms compete to provide the same services. ' Pastor Jim Henry, First Baptist Church of Orlando (personal communication, August 31, 2003). ' Hillebrandt, P. M. (1974). Economic Theory and The Construction Industry. London, Great BritainThe Macmillan Press, Ltd. (p. 37).

PAGE 17

Market Research Services of Florida conducted a study of 37 construction firms with $2 miUion to $215 million in construction volume and found that most CEO's do not truly believe they are in control of their growth.^ By reviewing growth histories, it was found that 70% were basically reactive marketers or dependent on the marketplace to dictate where, when, how, and if the company would grow. In specialty areas or markets of rapid growth, many reactive firms were successfiil until that point when the market or geographic area leveled off or declined.'* Construction organizations need a continuity of activity, not only to keep their resources fiiUy employed but to generate a return on investment that will attract the necessary capital for their continued existence and growth.^ Corporations seem to grow (increase in volume) or die. Few are able to remain at a constant level, no matter how hard they try. There is an inevitable need to expand, improve, or increase [a firm's] share of the market.^ Before a large D&C firm begins the process of getting in the door of a client, or even deciding on whose door to enter, a firm needs to know where the doors are. For the most part, operating companies are expected to find their ovra growth areas.^ The question then becomes where are the best future markets for a large D&C firm to ^ Pickar, R. L., AGC Construction Marketing Committee. (1995). A Contractor's Guide to Focus Sales and Increase Profitability: A Marketing Workbook for Contractors. Washington, D.C.: Associated General Contractors of America, (p. 69). * Pickar, R. L. (pp. 69-70). ^ Gerwick, B. C, & Woolery, J. C. (1983). Construction and Engineering Marketing for Major Project Services. New York: Wiley, (p. 2). * Gerwick, B. C, & Woolery, J. C. (p. 19). 'Hillebrandt, P. M., & Cannon, J. (1990). ne Modern Construction Firm. London: Macmillan. (p. 64).

PAGE 18

4 compete? The conventional wisdom is that D&C firms must go where the money is. The money is logically thought to be in the largest population centers and/or the markets closest to these population centers. In other words, start with a major city and target the perimeter counties. While there may be much truth and common sense to this approach, it is simplistic and not founded in proven theory of how a county's construction activity actually develops. While population and adjacency to major cities may be good rules of thumb, our research proposes that there may be better methodologies for strategically forecasting future market opportunities. Problem Statements Many questions must be answered by the marketing professional when tasked with finding new competitive markets for a D&C firm. Some of these questions may include; • Where are the best new markets for a large D&C firm? • What are the key indicators that best predict a potential market's construction activity? • What methodologies are available to identify these new markets? • Are complicated forecasting methods really more accurate than a simple, more direct, approach? • Which forecasting methods are the most accurate? • Finally, how should the various levels of market opportunity be segregated so the highest potential markets can be prioritized and pursued? Objectives of the Research A good hockey player plays where the puck is, a great hockey player plays where the puck is going to be. Retired hockey legend Wayne Gretzky^ BrainyMedia.com. Wayne Gretzky Quotes. Retrieved February 17, 2004, from http://www.brainyquote.eom/quotes/authors/w/wayne_gretzky.html.

PAGE 19

5 The overall objective of our research is to generate knowledge that will assist larger D&C firms with selecting the best future construction-market opportunities based on key construction indicators. While the objectives of oiu" research are to find answers to the strategic marketing research questions listed above, another important objective of our research is to develop and apply the logic of Newton's momentum theory to constructionmarket forecasting. The proposed momentum forecasting methodology will be used to convert key construction indicators and their associated levels of influence into a usable market-opportunity index for all of the counties within the State of Florida. Benefits and Significance of the Research From an academic perspective, the most significant contribution of our research is the identification and analysis of the key indicators of nonresidential construction activity. The analysis of these key indicators will provide a better understanding of which (and what type of) key indicators best predict a county's construction activity. The second most significant contribution of our research is applying the concept of momentum theory to construction-market forecasting. The proposed momentum forecasting approach was designed to be less complex but more meaningful than traditional statistical regression approaches. One of the differences of the momentum approach is that each variable includes three-dimensions. These dimensions include the relative size, rate of change, and influence of the variable. These three dimensions are combined into one new variable value (momentum). The momentum forecasting theory is intended to be a more understandable fi-amework that can be used to think about, and conceptualize, the complexities of construction-market forecasting.

PAGE 20

Another contribution is the appHcation and comparison of various methods of construction-market classification. Finally, the large firm, nonresidential market focus is another unique contribution of our research. From an industry perspective, the knowledge gained firom our research is useful to large D&C firms for several reasons. Their marketing resources could be more efficiently used to (a) identify developing competitive geographic markets such as individual counties, or county clusters or networks; (b) compare and prioritize these markets; and (c) better predict construction activity in these markets several years in advance. Organization of this Study Our research is divided into seven chapters (Figure 1-1). Chapter 1 includes a discussion of the issues leading to our research, statements of the problems to be researched, objectives of the research, and the significance of the research. The following paragraphs describe the content and organization of Chapters 2 through 7. Chapter \ Chapter 1 2 Introduction Literature Review \ \ \ Chapter IVIomentum IVIethodology \ Chapter Comparative Validation Methodologies Chapter 5 Key Indicators Findings Chapter Forecasting Methods Findings Chapter 7 Conclusions Figure 1-1. Organization of this study. Chapter 2 includes an overview of the management approaches to strategic construction marketing and is followed by a detailed discussion of the key indicators known to forecast the potential for construction activity. Finally, the various potential

PAGE 21

7 approaches to construction-market forecasting are reviewed. Several of these different forecasting approaches are selected for use in our research. Chapter 3 develops and applies the proposed momentum theory to strategic construction-market forecasting. Next, a discussion of what differentiates momentum analysis from other forecasting techniques is included. The research data sample and data sources are reviewed. Finally, the procedure for analyzing county momentum is detailed and a momentum index is derived. Chapter 4 discuses all of the steps taken during our research to comparatively validate the proposed momentum forecasting methodology to five alternative forecasting approaches selected from the literature review. A one, two and three year trend projection analysis is completed using all of the forecasting methods. The forecasted results from all of the techniques are comparatively validated and rank ordered against a county's actual construction activity. Finally, cluster regression analysis is used as a way to group the counties and is compared to two other classification techniques. Chapter 5 begins with a review of the overall results for the key indicator analysis, and is followed by a discussion of the specific findings for each key indicator variable. Next, results from the cluster analysis and the two other construction-market classification methodologies are reviewed. Finally, findings regarding the relationships between the key construction indicators and county clusters are presented. Chapter 6 begins with a review of the overall resuhs for the forecast methods including a discussion of the relevant statistics. This is followed by a review of the finding for each individual forecast methodology. The best forecasted fixture nonresidential construction-markets in the State of Florida are identified. Next, variance

PAGE 22

8 results between the projected county rankings and actual county rankings are presented. Finally, findings regarding the relationships between the forecasting methodologies and county clusters are reviewed. Chapter 7 presents what was found and learned by our research. This chapter begins with a review of the research intent and expectations and is followed by a summary of the research results. The overall threats to the research validity are discussed. Finally, several opportunities are identified regarding fiiture extensions and use of the presented research.

PAGE 23

CHAPTER 2 LITERATURE REVIEW This chapter provides a summary of the literature review completed for our research and is organized into three primary sections. The chapter begins with an overview of the management approaches to strategic construction marketing and planning. Next, key indicators of construction activity are identified, grouped and reviewed. Finally, potential methodological approaches to construction-market forecasting are reviewed and several methodologies are selected for use in our research. Management Approaches to Strategic Construction Marl^eting Strategic Marketing Definitions The words strategic and marketing have been given a variety of definitions throughout the literature. The following definitions are provided to give a common point of reference for this terminology. Strategic: Necessary to or important in the initiation, conduct, or completion of a strategic plan. Of great importance within an integrated whole or to a planned effect.' Marketing: The process of planning and executing the conception, pricing, promotion, and distribution of ideas, goods, and services to create exchanges that satisfy individual and organizational objectives.^ ' MerriamWebster Online Dictionary. Retrieved February 2, 2004, from http://www.m-w.com/cgibin/dictionary. ^ Bennett, P. D. Dictionary of Marketing Terms (2""' ed.). Lincolnwood, IL: NTC Publishing Group 1995 (p. 166). 9

PAGE 24

10 strategic Marketing Growth Models and Planning Processes The following sections of this chapter provide a summary of several business growth models and marketing plaiming processes that are specific to the design and construction industry and to our research. The intent of these sections is to show where our research fits into the marketing planning process. One of the most conventional views of marketing and strategic planning for the design and construction industry comes fi-om BNI Building News (2000). BNI begins by differentiating between sales and marketing. Sales is closing the deal, the specific project or program. It's the signing of a contract, the exchange of money for services. Everything up to that point is marketing.^ BNI differentiates between a marketing strategic plan and a marketing business plan. A strategic plan is a three to five year road map, whereas the marketing business plan is a one year increment of that plan.'* BNI then presents a five-step strategic plaiming process.^ These five general steps include; • Research and analysis of internal strengths and weaknesses and external opportunities and threats • Collecfive decision making • Organizational engineering and communication • Implementation of the plan • Evaluation and results Our research would be used during step one of BNI's strategic planning process. More specifically, our research focuses on the external opportunities of the large design ^ BNI Building News Society for Marketing Professional Services. (2000). Marketing Handbook for the Design & Construction Professional. Los Angeles: BNI Building News. (p. 13). " BNI Building News. (2000). (p.27). ' BNI Building News. (2000). (p. 28).

PAGE 25

11 & construction (D&C) firm. BNI describes the research and analysis of these external opportunities as The identification of external factors, determined by analyzing broad based social, demographic, cultural and economic trends for marketing and business implications, as well as specific possibilities presented by cUent/market requirements.^ Another conceptual framework for strategic marketing and planning comes fi-om Smyth (2000). Smyth presents four possible market-analysis viewpoints for the construction-marketplace.^ These market views are based on interrelationships among the origin of the view (internal or external) and the direction of the view (top down or bottom up). Our research emphasizes the top dovm market research view that is external to the organization. Smyth also presents a nine-step strategic planning process.^ These nine general steps have similar attributes to those presented previously in the BNI model and include • Business objectives • Marketing audit (the Four Views) • SWOT analysis • Assumptions • Marketing objectives and strategies • Estimate expected results • Identify alternative approaches • Implementation plan • Monitoring plan * BNI Building News. (2000). (p. 29). ' (Smyth, H., & NetLibrary Inc. (1999). Marketing and Selling Construction Services. Maiden MABlackwell Science, (p. 50)). * Smyth, H., & NetLibrary Inc. (1999). (p. 52).

PAGE 26

12 Our research would be used in the business objectives step of Smyth's planning process. More specifically, our study focuses on the business objectives relating to the geographic structure of the D&C firm. Smyth then outlines three traditional models for design and construction firm growth. These growth models are listed below. Our research would be used with Growth Model 2. • Model 1 . Expansion into existing markets • Model 2. Expansion into new markets • Models. Expansion by takeover or merger Another conceptual framework for strategic marketing and planning comes fi-om the Associated General Contractors of America (AGC 1995). Similar to BNI Building News, the AGC differentiates between a construction marketing approach and sales approach. The AGC's six step marketing approach^ is outlined below. Our research would be applied during the AGC's research and screening task. • Research and screen • Select target • Advertising, public relations, and networking • Trust • Sales • Close The AGC then outlines a ten-step marketing process'" similar to those discussed above. This process is designed to allow the construction firm to target particular markets and decision makers. The steps of this process are listed below. ' Associated General Contractors of America., & AGC Construction Marketing Committee. (1995). A Contractor's Guide to Focus Sales and Increase Profitability for the Associated General Contractors of America: A Marketing Workbook for Contractors. Washington, D.C.: Associated General Contractors of America, (p. 9). AGC. (1995). (p. 17).

PAGE 27

13 • Determine your mission • Set goals • Perform internal analysis • Perform external analysis • Establish marketing goals • Generate strategies • Research and refine strategies • Create and refine promotional and sales tactics • Implement the plan • Evaluation of results The AGC also provides a good explanation of an external market analysis: An external analysis examines the trends in the marketplace: hot-vs.-cold markets, local economic outlook, market types, available financing, and market needs. During the external analysis (in the up-and-down, cyclical construction arena) it is important to research basic factors that can create or eliminate a market place for a general contractor. These factors include the competition; the economic, social and political changes in the marketplace; and the need for particular infi-astructure, facilities, and contracfing services." The AGC then provides a list of the most common reasons that geographical expansion is initiated by a D&C firm.'^ These reasons are listed below. • A leveling off of the need in an existing market. • An increase in competition beyond the rate of area or market growth. • A major increase in price sensitivity. • There is a potential long-term downturn in the existing market. • The local area is historically very cyclical. • The local area is dependent on too limited a market for economic health. • There are strong management resources looking for an opportunity of personal growth in a more independent atmosphere. Another conceptual fi-amework for strategic marketing and planning comes fi-om Gerwick & Woolery (1983). First, Gerwick & Woolery recommend several ways for a "AGC. (1995). (p. 20). AGC. (1995). (p.60).

PAGE 28

14 construction firm to increase its contract volume.'^ These recommendations are similar to Smyth's growth models discussed above and are listed below. • Geographical expansion into new market areas. Greater market penetration or increasing the percent share of a firm's existing market. • Diversification of a firm's services on its own or through the acquisition of other firms. • Increase a firm's scope of services within a firm's area of expertise. The focus of our study is on the first recommendation, geographical expansion into new markets. Gerwick & Woolery outlined a six-step marketing plan.'"* The six primary steps of this approach are listed below. Establishing the firms long range objectives Evaluating alternative marketing plans Selecting a tentative plan for use Implementing the plan Monitoring the firm's performance while using this plan Revising the marketing plan as necessary to achieve the firm's desired objectives Our research would be used during step one of this six-step marketing plan. Still another conceptual fi-amework for strategic marketing and planning comes from Friedman (1984). Friedman begins by differentiating between construction-market forecasting and market planning. Planning is determining what a company wants in the fiitiire and developing methods to achieve it. Forecasting describes the type of external environment that can be expected.'^ Gerwick, B. C, & Woolery, J. C. (1983). Construction and Engineering Marketing for Major Project Services'. New York: Wiley, (pp. 38-39). j j "* Gerwick, B. C, & Woolery, J. C. (1983). (p. 38).

PAGE 29

15 Friedman presents a six-step sales process'^ for the construction industry similar to those previously presented. These steps are listed below. • Establishing corporate marketing objectives • Generating project leads • Qualifying prospects • Conducting sales interviews • Preparing proposals & presentations • Entering into contract negotiations and closing Our research would be used during step one of Friedman's process. Finally, a more complex systems approach to strategic construction marketing was presented by Fisher (1986).'^ Fisher's concept was based in part on the work of Adler 18 (1967). This systems approach embraces the marketing complexities of the organization and its interface with its environment through a series of input-output diagrams, interaction maps, data flow diagrams, and an overall marketing information system diagram.'^ While this approach offers an alternative way of looking at market planning, the models are; complex, time consuming to construct, and difficult to adapt to different situations; the models may also provide inaccurate results due to the lack of hard data; and finally they are perceived as too restrictive by marketing executives.^*^ " Friedman, W. (1984). Construction Marketing and Strategic Planning. New York: McGraw-Hill. (p. 12). Friedman, W. (1984). (p. 142). " Fisher, N. (1986). Marketing for the Construction Industry: A Practical Handbook for Consultants, Contractors, and other Professionals. London: Longman; J. Wiley. 18 19 20 Adler, L. (1967). Systems Approach to Marketing, Harvard Business Review. Sept./Oct. (pp. 105-1 18). Fisher, N. (1986). (p. 55-67, 1 15). Fisher, N. (1986). (p. 66).

PAGE 30

16 Fisher also presented a ten-step marketing planning approach. These ten steps are listed below. Our research would be used during steps 2 through 4. • Litemal company appraisal. • External company appraisal. • List existing, and identify new, business opportunities. • Assess future market potential for each opportunity identified. • Assess company ability to secure successful business in identified market sectors. • Rank options based on potential profit yield, prevailing conditions and key skills needed (existing or new). • Define and agree on marketing objectives. • Prepare detailed business plan for meeting objectives. • Putting the market plan into action. • Monitor and review plans in light of conditions encountered and performance achieved. Summary of Management Approaches There are probably as many ways to develop a strategic marketing plan as there are design & construction firms in the United States. What was learned fi-om the preceding literature review is that a D&C firm must systematically plan for what the firm wants to be in the future, and chart a course to achieve that vision. Strategic planning often yields new undertakings for the firm in the form of new geographic locations and offices, mergers and acquisitions, and start-ups of new businesses.'^^ While BNI, Smyth, AGC, Gerwick & Woolery, Friedman, and Fisher all offer a well structured set of tasks, conceptual models, and definitions in regards to sti-ategic construction-market planning, they stop short of offering a detailed methodology on how to actually implement the research, prioritize, and select a potential new constioictionmarket. Another observation is that the available construction marketing literature is ^' Fisher, N. (1986). (p. 133). ^ BNI Building News. (2000). (p. 33).

PAGE 31

17 focused on the sales aspects of constructing marketing for smaller firms, not the identification of new competitive markets for larger design and construction firms. This research is intended to be used as a top down appraisal tool for identifying new external business growth opportunities, and assessing the fiiture market potential for each opportunity identified. Our research is intended to be applied in the initial stages of a firms marketing planning process, during the tasks of research and screening. Our research is intended to initiate decisions regarding the geographical expansion and spatial structure of the organization. In short, the research goal is to locate the opportunities, not win an opportunity. Key Indicators of Construction Activity The second area of literature review for our research is the identification of key indicators of construction activity. The following sections of this chapter begin with a brief discussion of construction market segmentation. Next, a large number of potential key construction indicators are identified and reviewed. Finally, several of these indicators are selected for use in our research and are classified into groups to allow a more understandable discussion of their relationship to construction activity. Construction-Market Segmentation Berkowitz et al. (2000) defines market segmentation as the process of "aggregating prospective buyers into groups that (a) have common needs and (b) will respond similarly to a marketing action. The groups that result from this process are market segments, a relatively homogenous collection of prospective buyers."^'' BNI Building News {2QQQ>f^ provided a similar definition for market segmentation. " Berkowitz, E. N. et al. (2000). Marketing (6th ed.). Boston Massachusetts: Irwin McGraw-Hill. (2000). (p. 13).

PAGE 32

18 Thomsen (1989) classifies construction-markets in three primary segments. These segments include; (a) the construction-type markets (e.g., transportation, power, buildings), (b) the geographic type markets (e.g., U.S. market, Florida market, Orlando market), and (c) the service markets (e.g., architecture, engineering, construction, construction management).^^ Gerwick and Woolery (1983)^^ outline a similar market segmentation approach. Finkel (1997) provides a higher macro classification of the type of construction-markets. Finkel 's market segments include; (a) private residential, (b) private commercial (nonresidential), and (c) public construction. Engineering News Record (ENR) uses similar methods to segment construction-markets. ENR's eleven type of work classifications^^ are listed below. • Building • Manufacturing • Industrial • Petroleum • Water • Sewer & Waste • Transportation • Hazardous Waste • Power • Telecommunications • Other Smyth (1998) recognized the process of market segmentation is not as clear-cut as it appears. Public and private owners build both residential and nonresidential projects in ^* BNI Building News. (2000). (p. 17). Thomsen, C. (1989). Managing Brainpower: Organizing, Measuring Performance, and Selling in Architecture, Engineering, and Construction Management Companies. Washington, D.C.: American Institute of Architects Press, (p. 13). ^ Gerwick, B. C, & Woolery, J. C. (1983). (p. 20). Tulacz, G., & Powers, M. (May 19, 2003). The Top 400 Contractors. ENR. (p. 63).

PAGE 33

19 almost any geographic location. To better capture these relationships, Smyth developed a simple market segmentation model.^^ This model graphically shows the relationships between the traditional market segments in the context of the overall marketplace. Due to the cost and availability of research information, and to our research scope and schedule constraints, a broader more general market segmentation approach is taken for our research. Our research focuses on the public and private construction activity in the nonresidential construction-market at the Florida county level. Trends and Forces in the Marketplace Regardless of the market classification method chosen by a D&C firm, all of the market segments within the market place, and the market place itself, are shaped by changes in the marketing environment. These changes in the marketing environment are caused by uncontrollable trends and forces that are external to the D&C organization. Berkowitz et al. (2000) indicate that these environmental forces involve social, economic, technological, competitive, and regulatory changes.^^ These changes in the marketing environment are a source of opportunities and threats to be managed.^" The marketing literature uses the term environmental forces to describe the various changes within the marketplace. In the design and construction industry, the term environmental is more closely associated with the natural and physical sciences (i.e., natural environment and environmental sciences). To avoid any confusion, our research Smyth, H. J. (1998). Innovative Ways of Segmenting the Market: Practice Guide No. I. Oxford: Oxford Brookes University, Center for Construction Marketing, (p. 12). 29 Berkowitz, E. N. et al. (2000). (p. 13). Berkowitz, E. N. et al. (2000). (p. 74).

PAGE 34

20 uses the term key indicators, a term more conventional to the construction industry, to describe the forces and trends that are changing in the marketplace. Identiflcation of Key Indicators A review of the construction economics, construction marketing and macroeconomic literature has provided a wide variety of sources that identify the indicators that could potentially influence construction activity within a given market. The indicators found in this literature review are summarized in Appendix -A. Appendix A includes over 250 indicators in 56 different categories, the potential units of measurement, an example scale (or index), and potential data sources. Table 2-1 summarizes the quantity of indicators and indicator categories found in Appendix A. It should be noted that many of the variables (e.g.. Environmental and political variables) outlined in Appendix A are not used in our research for a variety of reasons. These reasons are outlined later in this chapter. The following paragraphs summarize several of the typical sources that were used to generate these key indicators. Table 2-1. Summary of construction activity variables listed in Appendix A. Variable categories Quantity of variables Economic, construction, and infrastructure 6 84 Community, government, and politics 8 82 Environmental 42 86 Total: 56 252 A comprehensive list of international construction indicator sources was published by the Organization for Economic Co-operation and Development (OECD).^' The publication includes a detailed list of the public and private sources of data for the main ^' Organisation for Economic Co-operation and Development. (2002). Main Economic Indicators: Comparative Methodological Analysis : Industry. Retail and Construction Indicators. Paris: Organisation for Economic Co-operation and Development, (p. 65-78).

PAGE 35

21 economic indicators of construction in 30 different countries. The OECD also lists the sources for the indicators of future and actual construction activity in the same 30 member nations. For its source of U.S. construction data, the OECD used the FW Dodge Corporation for future construction activity indicators, and the U.S. Census Bureau as the source for actual construction activity indicators. hi 1999, Standard & Poor's DRI (F.W. Dodge Division of McGraw-Hill Construction Information Group) published Building New Markets: Global Construction Market Opportunities and Risks?^ The purpose of this prospectus was to sell international construction-market research information services to large D&C firms. In Standard & Poor's Method and Analysis summary, the market forces affecting the construction-market in the short-term and the implications of long-run structural changes on construction-markets are discussed.^'' The following paragraph is a brief summary of the construction indicators discussed in the S&P DRI prospectus. In the short-term, the growth of GDP, interest rates, inflation, unemployment, international trade and financial linkages and exchange rates can affect constructionmarket demand. Other key indicators listed include domestic investment on commercial structures, dwellings, infrastructure and the country's ability to finance its projects. Long-run structural changes that influence a country's construction-market were subgrouped by Standard & Poor's into three broad dimensions. These dimensions include (a) shifts in the sectoral economic activity, (b) changes in the pattern of urbanization; and Standard & Poor's, DRI & F.W. Dodge. (1999, December). Building New markets: Global Construction Market Opportunities and Risks (Prospectus for Multi-Client Study). Lexington, MA: The McGraw-Hill Con^anies Construction Information Group. " Standard & Poor's, DRI & F.W. Dodge. (1999, December), (pp.1 1-13).

PAGE 36

22 (c) the demographic transition. Systematic sectoral shifts in economic activity were said to be the most influential and include movement from low to high incomes, and movement from agriculture through manufacturing into a service sector economy. The next most influential change on construction-markets is urbanization and includes a country's transition from widely dispersed small communities to large urbanized cities. As transportation services expand, secondary urban areas develop around the large cities. Demographic changes are the last most influential changes that affect a country's construction activity and include; population size, population growth, fertility and mortality rates, household size, education levels, and per capita income. The North American Construction Forecast (NACF) published a report in 2002 on the national U.S. construction activity outlook for the upcoming 2003 year.^'' In this report Ken Simonson, Chief Economist of the Associated General Contractors, was interviewed and said that there are two key indicators of U.S. construction health: construction employment and value put in place. Simonson went on to identify several key indicators in specific segments of the construction-market. These indicators are listed below. • Interest rates and unemployment for single family home construction. • Tax receipts, bond referendums and property values for the public construction sector. • Consumer spending on homes, automobiles and health related items for the broad private nonresidential segment, (e.g., building supply store construction, auto sales facilities, drug stores and health care facilities). Wright, R. (2002, October 16). U.S. Construction Activity Stagnant, but Promises Gradual Improvement; Government-Related Construction Shows Diminished Activity in 2003. North American Construction Forecast. Retrieved October 1, 2003, from http://www.nacf.com/simonson 02.html.

PAGE 37

23 • Unemployment, capacity utilization and profits for factory, office, warehouse, business related hotel, restaurant and car rental agency construction. Steele Analytics published Construction and Real Estate Market Pulse (2003)^^ on their website that lists eleven indicators for the U.S. construction and real estate market. These indicators are listed below. • Building permits • Construction employment • Construction equipment producer price index • Construction equipment shipments • Construction spending • Median sale price of existing single family homes • Existing single family homes sales • Lumber producer price index • Median sale price of new single family homes • New single family homes sales • Housing starts The New Jersey Construction ReporteP^ (a publication of the Division of Codes and Standards, New Jersey Department of Community Affairs) uses four major construction indicators to evaluate the activity of the New Jersey construction industry. These construction indicators include a quarterly comparison of; • Estimated cost of construction • Authorized housing units • Authorized office space • Authorized retail space The Metropolitan Washington Council of Governments published the report Economic Trends and Commercial Construction Indicators for Metropolitan Washington Construction and Real Estate Market Pulse. Steele Analytics. Retrieved October 1, 2003, from http://www.steeleanalytics.com/construction.htm. New Jersey Construction Reporter, March 2003 Highlights. Division of Codes and Standards, New Jersey Department of Community Affairs. Rehieved October 1 , 2003, from http://www.state.nj.us/dca/codes/cr/subfr)rm.shtml.

PAGE 38

24 (2003). This report identified twelve regional construction indicators that were used to forecast construction activity in the Washington D.C. area. These twelve regional indicators are listed below. • Population • Employment • Federal spending • Labor force • Construction • Mortgage rates • Home sales • Housing related inflation • Inflation • Income • Retail sales • Airline passengers Other sources for U.S. construction-market forecasting indicators and methodologies in use today are listed below. • Construction Review (U.S. Department of Commerce) • County Business Patterns (Bureau of the Census) • Survey of Current Business (U.S. Department of Commerce) • Bureau of Labor Statistics (U.S. Department of Labor) • Construction Online McGraw-Hill • Reed Construction Data • Associated General Contractors • Lend Lease Real Estate Investments • PricewaterhouseCoopers Environmental and Political Indicators There are as many reasons for including or excluding political and natural environmental indicators as there are variables in Appendix A. I have attempted to summarize three primary reasons why these variables have been avoided in our research. Economic Trends and Commercial Construction Indicators for Metropolitan Washington. Metropolitan Washington Council of Governments. Retrieved October 1, 2003, from http://www.mwcog.org/uploads/committee-documents/9FtYXw200307 1 5 144 1 12.ppt.

PAGE 39

25 These reasons include; (a) the problem of defining what the natural or political environment actually is, (b) the requirements for assessing the political and natural environment, and (c) the intended purpose of our research. First, the National Envirormiental Policy Act of 1970 (NEPA) mandates that federal government agencies assess the environmental impacts of actions "which may have an impact on man's environment."^^ Most state and local government agencies have adopted NEPA's regulatory framework for use at their levels. But the meaning of a man 's environment was not defined by this legislation nor has it been over 30 years later. Is the natural environment defined by wooded scenes, fresh air, clean water, low noise levels, and a pleasant suburban neighborhood? Or is it the dynamics of the food chain, endangered species, agricultural pesticides, and global warming? Is the political environment defined by the availability of health care, quantity of homeless persons, free child day care, and the percent of voters registered Democratic? Or is the political environment defined by public safety, the quality of public transportation, classroom student to teacher ratios and economic stability? These are all interrelated characteristics of a man's environment that may be impacted differently depending on the specific action taken. Further, all the variables listed in Appendix A can be made political. While it is generally agreed that the elements of the political and natural environment include the aesthetic, historic, cultural, economic and social aspects of a community, "the ultimate selection of what is really important in any one case is very much an art."^' National Environmental Policy Act, Title I, Sec 102(2) (A). Jain, R.K. et al. (2002). Environmental Assessment. New York: McGraw-Hill. (p. 5).

PAGE 40

26 Second, since our research is being conducted at a strategic level (i.e., county level), not at a project level, it is difficult to perform an actual assessment of the political and natural environment at the county level for several reasons. First, it is necessary to have a complete understanding and definition of the proposed construction action or project. Next, it is necessary to have a complete understanding of the surrounding environment being affected by the action caused by the location specific nature of political and environmental variables (e.g., air quality, water quality, noise levels, threatened species, public safety, health, social environment). Finally, the defined action or project would have to be combined with the setting to determine the interaction and changes that may occur. For an anti-growth or NIMBY organization to build up any type of resistance, they must have something to resist against. For these reasons, an assessment of the political and natural environment is normally conducted only at the project specific level. Key Indicator Selection and Constructs After the initial variables in Appendix A were identified, the preceding literature review was used to narrow the quantity of variables down to approximately 20 to 30. At this point, data availability became one of the primary determinants of which variables were finally selected. A total of sixteen key construction indicators were selected from the previous literature review and were used as independent variables in our research. Each of these sixteen variables including the variable name, type, unit, description, and source are outlined in Chapter 3. To simplify the discussion regarding the relationships between the key construction indicators, the indicators were grouped into six independent variable constincts. The independent variables and their associated constructs are shown in Table 2-2.

PAGE 41

27 Table 2-2. Key indicator constructs and associated variables. Population Geographic Initial Infrastructure Employment Economic Financial Advantage Transition Environment Resources Total Population ProximityDaily Vehicle Miles Construction Total Total Taxes Population Density Size Factor Traveled Payroll Employment Assessed Centerline Miles of Total Personal Gross Sales Commercial Roadway Income Price Level Land Value Road Density Average Wage Index Total Revenue Housing Starts These key indicator constructs are nothing more than logical groupings of the key construction activity indicators identified in the preceding literature review. The following paragraphs are descriptions of these six proposed constructs. Population Large increases in a county's total population drive the need for new buildings and infi-astructure. Multiple urban centers will further drive the need for connecting transportation and distribution infrastructure. It is hypothesized that total population, population change and density within a county are positively correlated with the county's construction activity. The variables that will be used to measure this construct include (a) a county's total population and (b) population density. Geographic advantage Secondary urban areas, or bedroom communities, typically develop around larger cities. This spillover effect contributes to the development of adjacent counties and is further amplified if the county is located between multiple large cities. It is hypothesized that a county's proximity to a larger urban center is positively correlated with the county's construction activity. The variable that will be used to measure a county's geographic advantage is a composite size/distance factor. This factor measures the relationship between the distance to an adjacent major city and the size of the adjacent major city.

PAGE 42

28 To calculate this factor, the distances between the county seat of Florida's 10 most populated counties and the county seat each of the 67 counties were obtained. These distances were multiplied by the population of the 10 largest counties. This provided a value that is weighted both by distance and population. This value was then multiplied by negative one (-1) and divided by 1,000 to transform the order and scale. A county's final proximity factor is equal to the sum of the values of the two closest and most populated counties. Initial infrastructure Public construction projects tend to become larger and more complex as a county develops. Population growth drives the demand for residential housing, water resources, and energy delivery systems. Unpaved roads are paved, and existing road capacity is increased. Counties will also encourage the development of industrial parks and foreign trade zones in an effort to lure manufacturing and service employment. This initial construction activity is a predecessor to larger public construction projects such as toll highways, power plants, and water and wastewater treatment facilities. It is hypothesized that growth in a county's initial housing and infrastructure is positively correlated with a county's nonresidential construction activity. The variables that will be used to measure this construct include (a) the total daily vehicle miles traveled on a county's roadways, (b) the total centerline miles of roadway, and (c) a county's road density. Housing starts, or the total annual value of residential construction permits issued in a county, will only be used in the direct forecasting methodology which is discussed later in this chapter and Chapter 4.

PAGE 43

29 Employment transition As a county develops over the long-term, employment typically moves from agriculture to manufacturing, then to the service sector. Shifts from the lower income employment sectors to the higher income sectors will stimulate investment in new facilities and infrastructure. This employment shift will also drive the need for a more highly educated work force. It is hypothesized that a shift from low to high-income employment sectors is positively correlated with a county's construction activity. The variables that will be used to measure a county's employment fransition include (a) total personal income, (b) average wages, (c) and total construction payroll. Economic environment A county's economic environment must be conducive to the growth of investment and employment. It is hypothesized that improvement of the economic environment within a county is positively correlated with the county's construction activity. The variables that will be used to measure a county's economic environment include (a) total employment, (b) gross sales, and (c) the Florida price level index. Financial resources A county must have the financial resources to construct new facilities and infrastructure. A county's bond rating is a reflection of its current financial position. Higher bond ratings enable a county to enter financial markets for essential borrowing at lower interest rates. It is hypothesized that an increase in a county's access to financial resources is positively correlated with a county's construction activity. The variables that will be used to measure a county's financial resources include (a) total tax collections, (b) taxable value of real property, and (c) a county's total revenue.

PAGE 44

30 Summary of the Key Indicators The various methodologies for segmenting the construction-market were reviewed. The approaches found in the literature review included segmentation by construction type, geographic location, and service type markets. They also included higher macro classifications such as private residential, private commercial (nonresidential), and public construction. All of the market segments within the market place, and the market place itself, are shaped by changes in the marketing environment. These changes in the marketing environment are caused by the uncontrollable trends and forces that are external to the D&C organization. These envirorunental forces involve social, economic, technological, competitive, and regulatory changes. Our research has used the term key indicators to describe the forces and trends that are changing in the marketplace. There appears to be a good deal of consistency in the literature regarding the key construction indicators at the macro level market segments of the industry (i.e., residential, commercial and public markets). But, which key indicators to use become more specific as a forecaster attempts to predict the activity within a major constructionmarket segment (e.g., housing, manufacturing, power, water supply, sewer/waste, industrial/petroleum, transportation, and telecommunications, environmental). An example of this is that a change in interest rates would have a greater effect on the housing market, than on the environmental construction-market. Opportunities for our research to add to the existing key indicator literature include a comparison of the indicators over multiple markets, the application of the indicators at a specific geographic level (not just by construction type or service), and a deeper

PAGE 45

31 understanding of the relationships between the key indicators and construction market activity. A total of sixteen key construction indicators were selected from the literature review to be used in the research. To simplify the discussion regarding the key construction indicators, the indicators were grouped into six independent variable constructs. These key indicators will be used to forecast the public and private construction activity in the nonresidential construction-market at the Florida county level. Approaches to Construction-Market Forecasting The final area of literature review for our research is a review of the approaches to construction-market forecasting. Forecasting is the linchpin of business because it is attempting to reduce risk and uncertainty. It is central to setting targets, budgets, hourly rates for professional services, planning of capital expenditure, and overhead recovery. ""^ The following sections of this chapter identify and describe the qualitative and quantitative forecasting techniques used during our research. Qualitative Techniques The Modem Forecaster written by Levenbach 8c Cleary (1984) state that the objective of qualitative forecasting techniques is to bring together in a logical, unbiased and systematic way all information and judgments that relate to the factors of interest.'*' Levenbach & Cleary present five qualitative forecasting techniques including; Delphi Fisher, N. (1986). (p. 123). Levenbach, H., & Cleary, J. P. (1984). The Modem Forecaster: The Forecasting Process through Data Analysis. Belmont, CA: Lifetime Learning Publications, (p. 15).

PAGE 46

32 Method, Market Research Focus Groups, Panel Consensus, Visionary Technology Forecasts Using Curve Fitting, and Historical Analogue.'*^ Berkowitz et al. (2000) identify and describe six additional qualitative forecasting techniques including; Direct Forecasting, Lost-Horse Forecasting, Survey of Buyers Intentions, Sales force Survey, Jury of Executive Opinion, and Survey of Experts.'*'' Direct Forecasting. Direct forecasting is the simplest of the qualitative forecasting methods listed above. "Probably 99.9 percent of all sales forecasts are judgments of the person who must act on the results of the forecast the individual decision maker."'*^ A direct forecast involves estimating the value of the forecast without any intervening steps. These estimates are opinions or judgments typically from experienced and competent executives inside a firm that know the market. They are gut feelings based on industry conventional wisdom or learned experience.'*^ Direct forecasting techniques are commonly used to forecast something about which the amount, type, and quality of historical data are limited. Examples of everyday direct forecasts include; should we bid on the advertised construction project? How much money should we budget for the bid proposal and presentation? How much time should be allowed to drive to the meeting? Direct forecasting will be used as the only qualitative methodology in our research. Although direct forecasting can be a quick and reasonably accurate prediction methodology, most knowledgeable executives still want some level of mathematical or statistical analysis performed to validate the direct forecast. "The objective is to avoid Levenbach, H., & Cleary, J. P. (1984). (pp. 15-17). Berkowitz, E. N. et al. (2000). (pp. 248-250). Berkowitz, E. N. et al. (2000). (p. 248). Berkowitz, E. N. et al. (2000). (p. 248).

PAGE 47

33 the use of only a single variable to represent a concept, and instead to use several variables as indicators, all representing differing facets of the concept to obtain a more well-rounded perspective."^^ The historical data required for our research is adequately available to allow a quantitative analysis. The following sections of this chapter will identify and discuss the quantitative methodologies available for construction-market forecasting. Quantitative Techniques Levenbach & Cleary present eleven commonly used quantitative forecasting techniques and further classifies them into either statistical (stochastic), or deterministic (causal) techniques."*^ The seven statistical techniques include Summary Statistics, Moving Average, Exponential Smoothing, Box-Jenkins (ARIMA), TCSI Decomposition, Trend Projections, and Regression Model. The four deterministic techniques include Econometric Model, Intention to Buy (Anticipation Survey), hiput-Output Model, and Leading Indicator. Montgomery (1976), Fisher (1986), Clapp (1987), Hair et al. (1998), and Berkowitz et al. (2000) all present similar quantitative techniques using somewhat different terminology. Levenbach & Cleary outline six factors that should be considered before deciding on the most appropriate projection technique. These six factors include; (a) characteristics of the data, (b) minimum data requirements, (c) time horizon to be forecast, (d) accuracy desired, (e) applicability, and (f) computer and related costs.''^ Hair, J. F. et al. (1998). Multivariate Data Analysis (5th ed). Upper Saddle River, New Jersey: Prentice Hall. (p. 10). Levenbach, H., & Cleary, J. P. (1984). (pp. 19-20). Levenbach, H., & Cleary, J. P. (1984). (pp. 25-32).

PAGE 48

34 Chambers et al. (1971) and Montgomery (1976) present similar selection factors using different terminology. Levenbach & Cleary go one-step further and developed a table to assist in the comparison and selection of the appropriate forecasting technique.''^ The application of Levenbach & Cleary' s table can quickly eliminate several quantitative techniques for our research. A time horizon of two years eliminates all of the statistical techniques except summary statistics, trend projections, and regression models. It also eliminates two of the four deterministic techniques leaving the econometric model, and input-output models. Brisbane Brown (1974), Roger Killingsworth (1990), and Standard & Poor's DRI (1999) have successfully applied econometric models to forecasting construction-markets and market costs. While there is widespread use of sophisticated econometric (causal) forecasts, there does not seem to be universal acceptance that econometric techniques produce consistently reliable and accurate forecasts (Armstrong, 1978; Granger and Newbold, 1977; Montgomery, 1976). Granger and Newbold contend that the econometric approach can be interpreted as a system in which a number of inputs are entered into a black box that transfers the values to an output. Montgomery states that the obvious limitation to the use of a causal model is the requirement that the independent variables must be known at the time the forecast is made. Another limitation of causal models is the large amount of computation and data compared with the time series model.^° Levenbach, H., & Cleary, J. P. (1984). (p. 28). Levenbach, H., & Cleary, J. P. (1984). (p. 28).

PAGE 49

35 Forecasting using a Summary Statistic technique was also evaluated for our research. While summary statistics are a good tool for generating a profile of the overall data, this technique may be less accurate than a multiple regression statistical model. The two remaining techniques, regression and trend projections are the most applicable to our research. Regression analysis Multivariate Data Analysis written by Joseph Hair et al. (1998) presents a decision tree to assist with the classification and selection of the proper multivariate-regression technique.^' Hair et al. outline three judgments the researcher must make about the research objective and the nature of the data:^^ (a) Can the variables be divided into independent and dependent classifications based on some theory? (b) If they can, how many variables are treated as dependent in a single analysis? (c) How are the variables, dependent and independent, measured (i.e., are they metric or non-metric)? There are two types of multivariate statistical analysis techniques used in our research, dependent and interdependent. These alternative regression techniques are used to estimate the level of construction activity in a county. These estimates are then compared to the results of the proposed momentum theory and to the actual known values. These dependent and interdependent regression techniques are discussed in the following paragraphs. Hair, J. F. et al. (1998). (pp. 20-21). "Hair, J. F. etal.(1998). (p. 18).

PAGE 50

36 Dependent regression techniques A dependent technique is one in which the prediction equation is dependent on, or estimated using, known dependent variable values. Our research involves one metric dependent variable (nonresidential construction activity) in a single relationship with multiple metric independent variables (key indicators). Therefore, the Hair et al. decision tree shows that either multiple regression or conjoint analysis could be used for this particular analysis. Conjoint analysis is a multivariate technique used specifically to understand how respondents develop preferences for products and services.^^ The most direct application is in new product or service development, allowing for the evaluation of complex products while maintaining a realistic decision context for the respondent.^'* Understanding this intended use does not fit our research goals; conjoint analysis was not selected for use. Regression analyses, and more specifically linear and multivariateregression analysis, have been chosen as the most appropriate statistical techniques for our research. Interdependent regression techniques An interdependent technique is one in which the prediction equation is not dependent on, or estimated using, known dependent variable values. It involves the simultaneous analysis of all the variables in the set. The goal is to find a way of consolidating the information contained in the original variables into an estimate or factor. This factor can then be objectively compared against the results of the momentum " Hair, J. F. et al. (1998). (p. 392). Hair, J. F. et al. (1998). (p. 15).

PAGE 51

37 forecasting methodology. Factor and cluster analysis are shown as the appropriate statistical techniques for our research in the Hair et al. decision tree. The variables used in the factor and cluster analysis techniques cannot be classified as either dependent or independent variables. In these techniques, "all of the variables are analyzed simultaneously in an effort to find the underlying data structure of the entire set of variables."^^ Factor analysis is the appropriate technique if the structure of the variables is to be analyzed. Cluster analysis is the appropriate technique if the cases (Florida counties) are to be grouped or classified. Trend analysis John M. Clapp has made several significant literature contributions to real estate market analysis and forecasting, hi his Handbook for Real Estate Market Analysis, Clapp presents a methodology for using regression for trend projection. Clapp's discussion also includes the limitations of the technique. For our research, a rolling one, two and three year trend projection analysis will be completed for the output variables of each of the forecasting methods. The purpose of these trend projections is to test the various methodologies for any inherent advantages relating to the duration of the forecast. The details of the trend forecast methodology used in our research are presented in Chapter 4. Gap analysis John Clapp has also made a number of contributions in the area of market and spatial gap analysis. "Market gap analysis determines whether there is (or will be) Hair, J. F. et al. (1998). (p. 22). Clapp, J. M. (1993). Dynamics of Office Markets. Washington, DC: The Urban Institute Press, (pp. 226227).

PAGE 52

38 unsatisfied demand in the entire market area."^^ Clapp outlines three methods for measuring market gaps: (a) gravity gap analysis, (b) expenditure-sales gap analysis, and (c) comparison of projected supply and projected demand. The first approach estimates the demand at a specific location. The last two approaches are used to determine the differences between supply and demand over a larger market area. The difference between the last two methods is that the second method uses the actual nimiber of units for the supply and demand projection, whereas the last method uses the percentage change. Because the data for the actual number of units is available, an adaptation of the second methodology (expenditure-sales gap analysis) will be used in our research. Our research uses gap analysis for several purposes. First, gap analysis will be used to compare a county's construction demand to its supply of construction resources. This will identify the Florida counties with construction demand that is larger or growing more rapidly, than the available supply. Second, gap analysis will be used as a tool to group the counties by their positive gap, balanced gap, or negative gap. A positive gap indicates that additional construction resources are needed in the coimty to meet the construction activity demand. Balanced gap indicates a balance between the supply of construction resources and the demand of construction activity. A negative gap indicates a surplus of construction resources exist for the corresponding demand for construction activity. Finally, gap analysis will be used to predict, and rank order, the level of construction activity in a county. The details of the different methodologies for applying gap analysis are presented in Chapter 4. Clapp, J. M.(1987). (p. 179).

PAGE 53

39 Law of Universal Gravitation In the 1680's, Sir Isaac Newton developed one of the most influential theories in Physics, the Law of Universal Gravitation. In 1929, William J. Reilly applied the logic of the law of universal gravitation to trade area analysis. Reilly's gravitational forecasting technique estimated the point between two cities where a customer would be equally likely to travel to one or the other.^^ Huff (1964) applied Reilly's gravitational model to estimating the visitation rates of customers from a given neighborhood to a given store.^^ Clapp (1987) improved and extended Reilly's work with the concepts of Gravity Capture Analysis,^*^ and Gravity Gap Analysis.^' Momentum analysis Sir Isaac Newton also developed the three laws of motion. These laws of motion were published in his book Principia^^ and remain as the cornerstone laws in the natural and physical sciences. Included in these three laws is the theory of linear momentum. The theory of linear momentum is used to mathematically relate the size and speed of an object or system. The logic of Sir Isaac Newton's momentum theory has been successfiilly applied in various areas of the social sciences and the natural sciences. Examples include; population momentum (Worid Bank Group 2003),^^ political campaign momentum Reilly, J. W. (1959). Methods for the Study of Retail Relationships. Austin, TX: University of Texas. Huff, D. L. (1964, July). Defining and Estimating a Trade Area. Journal of Marketing. Clapp, J. M. (1987). (pp. 173-178). Clapp, J. M. (1987). (pp. 180-186). *^ Newton, Isaac. (1687). Principia, Translated by Andrew Motte 1729. " Development Education Program Web, Glossary. The World Bank Group. Retrieved October 08, 2003, from http://www.worldbank.org/depweb/englisb/modules/glossary.html.

PAGE 54

40 research (Momentum Analysis Opinion Research, LLC 2003),^"* stock market momentum forecasting (Martin Pring 1993,^^ Hugh Clark 2002^^), and economic momentum ft! measurement (Missouri Economic Research and hiformation Center 2003). These last two applications are of the most interest to our research and are expanded upon below. Momentum forecasting in the stock market In the 1970's Dr. George Lane developed a securities market forecasting indicator known as the stochastic oscillator, or momentum indicator. Over the last 25 years Lane's forecasting methodology has been used and adapted by the financial industry to measure a securities rate of change. The Keystone Commodity Trading Guide (2001)^^ lists several type of common momentum indicators in use including: The Stochastic Oscillator, Rate of Change, Smoothed Rate of Change, Momentum Index, RSI, Williams %R, Commodity Channel Index, and the Moving Average Convergence/Divergence method. Other published methods include the Ford Value/Momentum model,^^ Momentum Analysis. Momentum Analysis Opinion Research, LLC. Retrieved October 08, 2003, from http://www.ballot.org/resources/Momentum_Analysis Brochure.doc. Pring, M. (1993). Martin Pring on Market Momentum. Sarasota, FL: International Institute for Economic Research, Inc. (p. 2); and Pring, M. (2002). Momentum Explained. New York: McGraw-Hill. Clark, H. (2002). Smart Momentum, the Future of Predictive Analysis in the Financial Markets. Chichester: John Wiley & Sons, LTD. (p. 7). Economic Indicators. Missouri Economic Research and Information Center. Retrieved October 13, 2003, from http://www.ded.state.mo.us/business/researchandplanning/indicators/momentum/index.shtnil. Momentum Indicators. Keystone Commodity Trading Guide. Retrieved October 13, 2003, from http://www.keystone-web.com/technicals/momentum.html. Value/Momentum Sector Analysis August 31, 2001. Ford Equity Research. Retrieved October 13, 2003, from http://www.fordequity.com/html/documents/Ss0801.pdf

PAGE 55

41 Trendcast System/" Stochastic Momentum Index/' Relative Momentum Index/^ and the CBR Stock Market Momentum Indicator/^ While there are many differences between these methodologies, these momentum indicators generally measure the price of a security relative to the high/low range over a set period of time. Momentum investing is based on the idea that stocks which have performed well over some interval in the past will tend to perform well in the future/'* This same logic is applied to a county's level of construction activity over time in our research. Economic momentum The Missouri Economic Research and Information Center (MERIC) publishes an annual Index of State and Economic Momentum. This index, developed by the late State Policy Reports editor Hal Hovey, is a composite of percentage changes in personal income, population, and employment at the county level. The index measures momentum in a county relative to the overall economic momentum of the state. "An index equal to 0 means the county realized average economic growth during the decade. Answers to Frequently Asked Questions. Trendcast, LLC. Retrieved October 13, 2003, from http://www.trendcast.com/amazing/faq.htni. " Stochastic Momentum Index. Paritech Inc. Retrieved October 13, 2003, from http://www.paritech.com/education/technical/indicators/momentum/stochasticl.asp. TradingSolutions Function Library, Relative Momentum Index [RMI]. Trading Solutions. Retrieved October 13, 2003, from http://www.tradingsolutions.com/functions/RelativeMomentumIndex.html. Stock Market Momentum Indicator. Commodity Research Bureau. Retrieved October 13, 2003, from http://www.crbtrader.com/crbindex/nsmmi.asp. Capeci, J. D. & Campillo, M.. (April 2002). Global Sector Momentum in the Emerging Markets. Cambridge, MA: Arrow Street Capital, (p. 2).

PAGE 56

42 An index less than zero indicate relatively sluggish growth, while an index greater than zero indicates relatively prosperous growth."^^ Hovey improved the original economic momentum index by adding the measures of county economic share and influence. MERIC defines economic share as the percentage of the state's economy that is accounted for by an individual county. The economic share is measured as the average of the percentage of the state's employment, population, and personal income that occurs in a particular county. MERIC defines economic influence as the product of the momentum index and economic share score, and is calculated by multiplying the two indices. Thus, a county with a high momentum score and a large economic share has a large level of economic influence, and is considered an important economic driver for the state. MERIC has also applied this same methodology to the regional and state geographic level. The methodology for measuring economic momentum is the same for these geographic levels. While these gravitational and momentum methods may seem unrelated to the construction-market forecasting focus of our research, the significance of Reilly, Huff, Clapp, Lane and Hovey's research is that they have demonstrated how the fi-amework of natural science theories can be successfully adapted and applied to social science research. Summary of Forecasting Approaches No single model can be considered universally adaptable to any given forecasting situation. Thus a basic principal is to utilize more than one projection technique. The Economic Indicators. Missouri Economic Research and Information Center. (2003).

PAGE 57

43 purpose of using more than one methodology is to insure that the forecasting approach will be as flexible as possible and that the forecaster's judgment is not overly dependent 76 on one projection technique. Opportunities for our research to add to the existing literature include the development of an accurate forecasting technique that is more understandable than complex regression approaches but more meaningful than simple direct approaches. Another opportunity includes developing a technique that integrates the variable dimensions of size, rate of change, and influence. Finally, the literature appears to be missing the application and comparison of multiple forecasting methodologies over several diverse markets. Our research uses six different techniques identified in the previous literature review to forecast a county's construction activity. These six methodologies include; 1), direct forecasting, 2) multivariate-regression, 3) factor analysis, 4) gap analysis, 5) MERIC economic momentum analysis and 6) momentum analysis. The following chapter will discuss how momentum theory was derived and apply it to strategic construction-market forecasting. In Chapter 4, the estimates from all six of these methodologies will be compared against the actual values of nonresidential construction activity for all of the counties in the state of Florida. Summary of Research Questions • Which, and what type of, key indicators best predict a county's nonresidential construction activity? • What are the relationships between these key indicators and county construction activity? Levenbach, Hans & Cleary, James P. (p. 35).

PAGE 58

44 Can a more accurate method of forecasting construction activity be developed and applied such as the proposed momentum methodology? Are complicated forecasting methods really more accurate than simple, more direct, approaches? Can the various construction-markets be segregated using methodologies such as gap or cluster analysis so the highest potential markets can be prioritized and pursued? Where are the best new markets for a large D&C firm in the state of Florida?

PAGE 59

CHAPTER 3 MOMENTUM THEORY DERIVATION AND APPLICATION Make things as simple as possible, but not simpler.' -Albert Einstein responding to the question of how he explained the difficult theory of relativity with such a brief equation as E=MC^. The philosophy behind Einstein's statement was used as the premise behind the development of the momentum forecasting approach. As discussed in Chapter 2, the momentum analysis approach was designed to be less complex but more meaningful than traditional statistical regression approaches. This chapter discusses all of the steps taken during our research to derive and apply the methodology of momentum theory to strategic construction-market forecasting. The chapter begins with an introduction to the momentum forecasting theory. The differences between this momentum methodology and multivariate statistical regression approaches are discussed. This is followed by a brief overview of the data sources and sample. Finally, the procedure for applying and analyzing county momentum is then detailed, and a momentum index is derived. In the next chapter of our research, the results of this momentum analysis are then comparatively validated against five alternative forecasting methods. Momentum Theory Introduction Momentum is a commonly used term to describe objects in motion. The concept of momentum is used to relate the size and speed of an object or system. Examples include; ' Brallier, J. (2002). Who fVas Albert Einstein? New York, Grosset & Dunlap. (p. 47). 45

PAGE 60

46 a fast train has more momentum than a slow train; the football team has lost its momentum going into the fourth quarter; or the latest government reports show that the economy may be gaining momentum. As discussed in Chapter 2, the theory of linear momentum was derived from Sir Isaac Newton's Three Laws of Motion. The linear momentum of an object is defined as a product of its mass and its velocity.^ This momentum theory is mathematically shown in Equation 3-1 . The magnitude of momentum "p" at any moment is equal to the numerical product of the mass "m" times the velocity "v". (Eq. 3-1) p mv The importance of the theory of momentum for our research is that it is a conserved quantity. The Law of Conservation of Momentum is stated as follows; The total momentum of an isolated system of bodies remains constant.^ Although the momentum of each of the objects changes as a result of the collision, the sum of their momentum is found to be the same before and after the collision. A system is defined as a set of objects that interact with each other. The total momentum of a system "Ptotai" is equal to the sum of the momentum in the components "Pn" of the system. This conservation of momentum theory is mathematically shown in Equation 3-2. (Eq. 3-2) Ptotai = Pi +P2... +Pn or Ptotai = miVi + m2V2... + mnVn Newton's theory assumed that the external forces (i.e., fiiction) on the isolated system of objects under study are zero. In our research, different key indicators (objects) ^ Giancoli, D. C. (1980). Physics. Englewood Cliffs, NJ: Prentice-Hall, Inc. (p. 1 16). ^ Giancoli, D. C. (1980). (p. 120).

PAGE 61

47 will influence a county's construction activity at different levels. For example, it is likely that a county's construction activity is more influenced by a change in its total population than by a change in the education level of its population. For this reason, the influence variable " i " will be added to the calculation of momentum in a county. The final equation used to calculate the total momentum in a county is shown in Equation 3-3. (Eq. 3-3) Ptotai = Pi +P2... +Pn or P,otai = miVi 1 1 + maVi 12... + ninVn in Differentiation between Momentum and Regression Approaches There are two primary differences between the proposed momentum model and traditional regression approaches. The following paragraphs are a summary of the two most significant differences including; (a) variable dimensions, and (b) multivariate measurement. First, traditional multivariate-regression analysis equations are constructed with a linear combination of one-dimensional variables with empirically determined weights. The variables are specified by the researcher, and the weights are determined by the regression technique. The result is a single output value representing a combination of the entire set of variables that best achieves the objective of the specific multivariateregression analysis. A typical regression variate can be stated mathematically as shown in Equation 3-4. (Eq. 3-4) Variate output value (Y) = |8o + jS, X, + 182X2 + ftXj + . . . + jSnXn The momentum analysis equation is also constructed with a linear combination of variables that are specified by the researcher. The difference in the momentum approach is that each variable includes three dimensions. These dimensions include the relative size, rate of change, and empirically determined influence of the variable. These three dimensions are combined into one new variable value (i.e., momentum). This new

PAGE 62

48 variable value provides a more meaningful representation of the original variable and the relationships within, and between, other variables. The momentum variate can be stated mathematically as shown in Equation 3-5. (Eq. 3-5) Variate output value (Ptotai) = niivi i i + m2V2 i 2 • • + ninVn In or Ptotai = P 1 +P2... +Pn The second fundamental difference between the techniques is that the output of the momentum equation is really a multivariate measurement or summated scale, not an estimate of the true dependent variable value. As with regression analysis, the result of the momentum equation is a single value representing a combination of the entire set of variables. But in the case of momentum, the variate value is derived without using statistical regression. The momentum variate value is a summation of the variables joined together as a composite measure of the concept under study (i.e., construction activity). These two different variate measurement techniques are reconciled in the research when the estimated variate values are calculated for each case, and the cases (counties) are rank ordered and compared. Data Sample and Data Sources Historical data was collected for all of Florida's 67 counties for use in our research. A map of the State of Florida's 67 counties is shown in Figure 3-1. A thirteen year time period was used for our study that started in 1990 and ended in 2002. Our research involves the analysis of one single dependent variable and its relationships to sixteen independent variables. This resulted in a data set of over 14,800 values (67 counties x 13 years x 17 variables). The data for our research was collected from a variety of secondary sources. Missing or erroneous data was corrected by using mean or trend

PAGE 63

49 projection techniques. Table 3-1 outlines each of the variables including the variable name, type, unit, description, and source. The actual variable data for each of the thirteen years in the study is included in Appendix B. Due to data license agreement restrictions, the nonresidential permit values for the years 1996 through 2002 are not included. Other data required to replicate our research is available by contacting the author. Momentum Theory Applied to Strategic Construction-Market Forecasting Applying the logic of momentum theory to strategic construction-market forecasting is fairly straightforward. The momentum theory variables introduced above have been redefined for the purposes of our research and are shown in Table 3-2. The measurement of momentum in each county is relative to the other 67 counties in Florida. A county with high construction momentum (p) would be characterized as being a large existing market (m), with rapid positive change (v) in the key variables that most influence (1) construction activity. Conversely, a county with low construction momentum (p) would be characterized as being a small existing market (m), with slow or negative change (v) in the key variables that most influence (i) construction activity. Methodology for Analyzing County Momentum The following sections of this chapter detail the methodology for calculating a county's momentum and the corresponding momentum index and slope. The seven general steps for the momentum analysis are outlined in Figure 3-2.

PAGE 64

50 Figure 3-1. Map of the 67 counties in the State of Florida.

PAGE 65

51 Table 3-1. Research variable data, descriptions, and sources. No. Variable Code Variable Name Type Unit Description Source Dependent Variable: 1 NRPERMIT Nonresidential Metric USDS Total annual value of 1) Building Permit Activity in Florida 1990 to Construction Ratio (1,000's) nonresidential construction 2003, University of Florida Bureau of Permits permits issued per county. Economic and Business Research. 2) Annual Construction Starts Data 1996-2002. McGrawMitt Construction Dodge. Independent Variables: 1 POP Total Metric Person The annual computed number Florida Statistical Abstracts 1990 to 2003, Population Ratio each of persons living in a county University of Florida Bureau of Economic and (1,000's) (resident population). Business Research. 2 PDENSITY Population Metric Persons 1 he annual computed number Density Ratio (1,000's) of persons living in a county University of Florida Bureau of Economic and per S
PAGE 66

52 Table 3-2. Comparison of momentum theory variable definitions. Variable Newton's momentum theory definition. Construction-market momentum definition. Ptotal Total momentum of a system. Total momentum in a construction-market (county). P Momentum of a component of the system. Momentum of a key variable. m Mass of a component of the system. Mass (size) of a key variable. V Velocity of a component of the system. Velocity (percent change) of a key variable. r I Analogous to an external force on an object or isolated system. Influence (strength) of a key variable. 1 Calculate variable velocity Calculate variable influence "i" Calculate variable mass L-t/ Figure 3-2. Seven general steps of momentum analysis Calculate variable momentum | "p" Calc total county momentum "Plot" Calculate momentum index "oindex" Calculate momentum index slope | "oslope" Mass of key indicators The first step is to calculate the mass "m" of the key variables. Mass is calculated using the actual value of the key variable. The variable value is normalized and is expressed in terms of its proportion to the range of values for the same variable. A value of one (1) has been added to the equation to eliminate a "0" effect during multiplication in the final momentum calculation. The range of output values for this calculation is 1.0 to 2.0. This mass calculation is shown in Equation 3-6. (Eq.3-6) Xn-Xn ^max " -^min hi Equation 3-6, "m" equals the calculated mass of the variable, Xn equals the variable's actual value, Xmin equals the minimum actual value of the same variables, and Xmax equals the maximum actual value of the same variables.

PAGE 67

53 Example mass calculation: If the variable has minimum and maximum values of 200 to 300 respectively, and a particular county has a value of 240, the mass for that variable would equal 1 .40. Velocity of key indicators The next step is to calculate the velocity "v" of the key variables. The variable value will be expressed in terms similar to the previous mass (m) calculation. The only difference is that velocity is measuring the aimual percent change of the variable, not its size. The velocity calculation is shown in Equation 3-7. The variable value is normalized and is expressed in terms of its proportion to the range of values for the same variable. A value of one (1) has been added to Equation 3-7 to eliminate a "0" effect during multiplication in the final momentum calculation. The range of output values for this calculation is 1 .0 to 2.0. (Eq.3-7) 2kziXmin hi Equation 3-7, "v" equals the calculated velocity of the variable, Xn equals the variables armual percent change, Xmin equals the minimum armual percent change of the same variables, and Xmax equals the maximum annual percent change of the same variables. The annual percent change of a variable was calculated using the example Equation 3-8. (Eq. 3-8) Annual percent change = (Year 200 1 Year 2000 / Year 2000) Example velocity calculation: If the variable has minimum and maximum values of 2.0% to 3.0% respectively, and a particular county has a value of 2.4%, the velocity for that variable would equal 1 .40.

PAGE 68

54 Influence of key indicators The next step is to calculate the influence (i) of a key variable on a county's total construction activity. The Pearson Correlation Coefficient is a statistical measure that has been used as the value for influence in our research. The Pearson correlation coefficient is a measure of linear association between two variables. Values of the correlation coefficient range fi-om -1 to 1. The sign of the coefficient indicates the direction of the relationship, and its absolute value indicates the strength, with larger absolute values indicating stronger relationships. The statistical software Statistical Package for the Social Sciences (SPSS) Base version 11. 5 (2002) was used to calculate the Pearson correlation coefficient on the dataset. This influence calculation is shown in Equation 3-9. A value of one (1) has been added to the equation to eliminate a "0" effect during multiplication in the final momentum calculation. The range of output values for this calculation is 1.0 to 2.0. (Eq. 3-9) i = IPI + 1 in Equation 3-9, " i " equals the calculated influence of the variable, and " |/3| " is the absolute value of the Pearson correlation coefficient. The Pearson coefficient was computed for all sixteen independent variables included in the data sample during the years 1990 through 2002. A two-tailed test of significance was used. In comparison to the other variables, the AVEWAGE, PLINDEX, PDENSITY, RDENSITY and PROX independent variables consistently demonstrated weaker correlation (< 0.750) to the dependent variable nonresidential permits (NRPERMIT). These five independent variables were dropped and are not included in the analysis of county momentum or the other forecasting methods.

PAGE 69

55 In addition to these five excluded variables, the independent variable RPERMIT will be used in our research only as a variable in the direct forecast methodology. The RPERMIT variable will be excluded fi-om the momentum analysis and all other methods in our research. Out of the sixteen variables originally proposed in Chapter 2, the ten remaining variables will be used as the key indicators for all of the forecasting techniques in our research. The results for the Pearson's correlation calculation are presented in Table 3-3 along with their minimum, maximum and average correlation. The variables are sorted by their average correlation in descending order. The five excluded variables have been shown below the dashed line for reference purposes. A detailed discussion of each variable's correlation to NRPERMIT is included in Chapter 5. Momentum of key indicators The next step is to calculate the momentum (p) for each key variable. This momentum calculation is shown in Equation 3-10. The momentum for a particular variable (pn) is the product of the key variable's mass (m), velocity (v) and influence ((). The maximum momentum value for a variable equals 8 (2 x 2 x 2), and the minimum momentum value for a variable equals 1(1x1x1). (Eq. 3-10) Pn = mnVn(n A variable's momentum can also be thought of as a three-dimensional or volumetric measurement. This three-dimensional relationship is shown in Figure 3-3(a). Total momentum of a county The final step is to calculate the total momentum (Ptotai) in a county. As discussed eariier in this chapter, total momentum is a conserved quantity that is equal to the sum of the momentum in the components (variables) of the system (county). This relationship is

PAGE 70

56 shown in Figure 3-3(b). Using Equation 3-2, the total momentum in a particular county is the sum of the momentum for each key variable within the county. The results of the momentum calculation for each year of the study are shown in Table 3-4. The 67 counties were sorted by their average momentum in descending order. (a) Variable momentum (pn) (b) Total momentum (Ptotai) Figure 3-3. Three-dimensional relationships of momentum, (a) Variable momentum, (b) Total momentum. Derivation of the Momentum Index The construction-market momentum index is a measure of the cumulative value of a county's total momentum. The base year for the index is 1990 with an initial value of 100. The calculated value of a county's total momentum was divided by 10 for scale. An example equation for calculating the momentum index (OINDEX) for a county in a given year is shown in Equation 3-11. (Eq. 3-1 1) OINDEXi99i = P1990 + (P1991 / 10) For example; if a county's 1990 momentum index value was 100, and its 1991 momentum value was 35, the county's 1991 momentum index value is 103.5 (i.e., 100 + (35/10) = 103.5).

PAGE 71

57 00 < X O O n 00 ON /-) ON ON O NO On oo NO On 00 ON CO On On 00 in 00 in o ON d d d d d d d d d d d o p NO ON ON ON NO *o On m ON 00 ?N ON ON NO ON NO ON >n On o 00 d d d d d d d d d d d o o o NO On 00 On On NO o> o NO On >n On o 0\ CM 0\ 00 fN On On On NO i—< ON o oo d d d d d d d d d d d o o o NO 00 00 oo NO
PAGE 72

58 The momentum index for each of the 67 counties was calculated and plotted from 1991 through 2002. A rolling one, two and three year trend projection analysis was completed for the momentum index. This trend projection analysis methodology is described in Chapter 4. The results of the momentum index calculations are included in Table 3-5. The 67 counties were sorted by the 2002 momentum index in descending order. This momentum index will be used as an indicator to forecast a county's nonresidential construction activity (NRPERMIT). Momentum Index Slope The results from the preceding momentum index calculations were plotted and are shown in Figure 3-4. This graph shows the momentum index for all 67 counties from 1991 through 2002. When observing this graph it can be seen that the lines of the counties with the highest momentum indexes have a greater line slope while the lines of the counties with the lowest momentum indexes have a lower slope. This slope will be used as the second indicator from this momentum methodology to forecast a county's nonresidential construction activity. The slope of a county's index in a specific year was measured by the slope of the linear regression line through the momentum index data points for the preceding five years. The slope is the vertical distance divided by the horizontal distance between any two points on the line, which is the rate of change along the regression line. The momentum index slope for each of the 67 counties was calculated. A rolling one, two and three year trend projection analysis was completed for the momentum slope. This trend projection analysis methodology is described in Chapter 4. The resuUs of the momentum index slope calculation are discussed in Chapter 6.

PAGE 73

59 Summary of Momentum Theory The intent of this chapter was to introduce a new forecasting methodology based on Sir Isaac Newton's natural science theory of momentum. This unique new methodology was applied to strategic construction-market forecasting. The two most significant differences between this new momentum methodology and traditional statistical regression approaches were discussed and include; (a) variable dimensions, and (b) multivariate measurement. The data sample and data sources used in our research were reviewed. Finally, the procedure for applying and analyzing county momentum was detailed. This procedure included; calculating the mass, velocity and influence of the key indicators; calculating the key indicator momentum and total county momentum; and calculating the momentum index and its corresponding slope. hi the next chapter of our research, the results of this momentum analysis are comparatively validated against five alternative forecasting methods. These alternative methodologies were previously outlined in Chapter 2.

PAGE 74

60 Table 3-4. Total momentum of a county for years 1991 through 2002. County" 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Average Dsde 32.84 50.73 57.08 52.09 30.49 54.46 57.41 57.59 57.54 57.56 56.82 53.38 54.83 Broward 45.43 45.80 47 34 47.14 43.33 47.84 50.24 49.07 50.99 51.22 50.79 47.87 48.09 Orange 41.85 40.37 41J0 37.98 40.00 45.25 47.51 46.24 47.11 46.59 44.75 43.24 43J3 Palm Beach 40.76 40.31 42.17 40.46 39.15 43.73 45.36 45.42 4454 46.62 44.41 43.63 43.08 HtUsborough 39.77 40.03 39.47 40.52 38.20 43J4 45.29 43.85 46.87 43.29 43.83 44.43 42J9 Duval 37.87 3733 39.61 37.08 37.49 40J4 44.54 41.01 41.19 40.12 40.15 39.45 39.70 Fjnellas 37.62 36.44 3736 36.72 34.82 38.65 41.19 39.53 39.50 40.45 38.99 37.87 3826 Ltt 34.03 31.62 34.76 34.28 32.34 3434 37.25 33.82 37.27 35.92 35.96 36.77 3503 Polk 32.75 31.63 31.80 32.66 29.47 33.63 36.29 34.61 35.28 33.72 32.26 34.12 3333 Brevard 34.10 33J3 3225 31.77 29.17 33.43 33.25 32.32 34.75 35.67 33.33 32.43 33.19 Seminole 31.43 31.53 32.06 29.96 29.43 33.12 36.20 33.65 36.43 32.97 32.64 34.57 33.00 Collier 32.41 30.40 3332 30.94 29.76 32.81 33.62 35.18 34.76 33.74 33.78 31.60 32.86 Voluaa 33.01 31.01 31.83 30.92 29.91 32.64 34.32 31.92 34.14 33.68 32.23 3219 3233 Sarasota 31.13 30.45 30.84 31.45 28.75 3236 33.12 32.98 33.58 32.62 33.24 32.29 32.07 Mali GO 30.30 30.99 30.83 31.51 28.51 31.66 34.78 33.63 34.40 31.07 30.65 31.73 31jS7 Manatee 31.77 29.94 30.11 31.49 29.22 30.72 32.90 30.27 33.27 31.49 31.13 31.96 31.19 Osceola 29.53 31.17 30.73 29.52 28.03 3132 31.84 34.06 33.96 30.54 30.06 32.37 31.10 Pasco 32.36 29.30 29J3 29.39 27.54 31.83 33.37 30.82 33.46 30.59 31.54 32.54 31.03 Uke 29.44 30.06 30.27 29.83 2837 31.11 33.64 31.79 33.66 30.66 30.71 30.71 30.85 Escambia 29.98 30.86 30.45 30.37 27.74 3152 33.05 31.00 3236 30.32 29.72 31.04 30.73 Saint Johns 29.43 29.31 29.63 29.69 28.89 29J4 34.61 31.84 32.34 31.01 29.92 31.94 30JS8 Leon 30.89 29.98 30.14 30.33 28.42 31.14 31.74 31.21 32.39 29.39 29.63 30.72 30J0 Walton 30.32 29.56 28J4 28.43 28.13 30.90 32.36 31.22 32.18 29.47 30.41 31.29 3023 Alachua 29.66 30.14 29.74 29.49 27.81 30.72 31.47 30.84 31.53 29.42 30.86 29.85 30.13 Okaloosa 31.14 29.45 31.03 28.99 29.38 30J6 30.29 31.93 30.46 29.39 29.37 29.22 30.10 Santa Rosa 31.30 31.53 2933 29.68 27.42 31.10 32.15 30.33 28.92 29.37 29.86 2930 30.03 Hernando 28.49 29.97 29J26 31.39 27.27 29J3 33.59 29.99 29.52 29.03 2934 30.15 29.80 Flagler 29.21 27.87 30.88 30.48 28.92 30.13 31.32 29.68 31.78 29.30 28.95 28.81 29.78 Charlotte 31.16 28.33 29.07 28.66 26.81 29.67 31.37 30.44 3057 29.61 30.72 30.08 29.74 Saint Lode 31.36 28.24 2852 29.55 27.84 2930 30 39 29.70 30.69 30.08 29.68 30.44 2965 aay 28.57 28.01 29J59 28.95 26.91 30.05 31.17 30.81 30.86 29.47 29.00 31.14 29J5 Wakulla 29.54 28.65 28J3 29.46 27.20 32.03 33 32 29.55 30.71 28.45 29.64 27.80 29J5 atau 31.37 28.69 29.64 29.12 25.39 29.41 30.02 30.50 31.02 29.91 29.08 30.28 29J4 28.81 27.07 27 /S5 27.30 27.02 30.61 32.91 31.26 32.93 29.03 29.23 29.17 29.42 BV 29.13 29.83 28J7 28.39 27.92 30.11 31.28 29.34 29.46 28.02 28.92 28.49 29.12 Ifirtfai 27.90 27.32 2938 28.06 27.64 29.61 31.38 29.87 29.91 30.20 28.37 29.45 29.09 Nassau 29.87 29.02 3037 25.19 26.05 29.27 31.14 28.36 30.35 30.46 27.77 31.02 29.07 filfian River 27.91 27.65 2653 29.62 28.03 29.61 31.65 30.14 29.85 27.08 30.26 28.22 2851 Oilchiist 31.07 25.92 29J54 28.64 25.26 30.83 30.30 29.07 28.82 27.91 29.29 27 39 28JS8 Washington 30.20 26.88 26.77 29.20 30.30 29J7 29.73 27.46 31.29 25.77 28.96 28.13 28JS6 Columbia 27.91 27.80 29.41 29.52 2731 30.03 31.13 28.23 29.00 27.94 27.20 28.11 28JS3 Levy 28.46 28.66 28.46 28.72 23.68 30.00 30.25 28.70 29.28 27.58 28.75 2828 28.57 Putnam 28.74 28.99 yoso 27.29 27.94 26.49 28.22 27.92 30.24 27.34 28.70 28.48 28.40 Monroe 27.00 28.17 30.09 27.00 26.84 28.24 29.50 29.82 29.13 23.76 29.18 27.75 2821 Highlands 29.87 29.04 28.05 28.32 23.66 27.90 29.75 27.68 29.79 27.04 27.39 27.78 28.19 Hendry 30.64 28.67 26.48 28.40 26.54 27.40 30.46 27.45 29.28 27.48 25.36 27.45 2757 Franklin 29.40 27.54 28.84 28.69 28.49 24.46 29.79 26.02 28.76 27.13 28.86 26.93 2751 Suwannee 26.76 27J2 29.16 28.28 25.99 28.87 27.76 28.96 28.74 27.31 2751 27.48 2750 Dixie 24.64 29 J4 30.03 28.42 24.06 27 J6 28.20 26.94 28.46 28.98 29.05 28.68 27.88 Baker 27.72 27.48 26.13 27.83 25.47 27.88 30.17 27.96 28.84 28.56 27.88 28.39 27.86 Ufsyette 27.00 26.80 23J9 26.85 27.64 28 J 1 28.43 27.77 31.19 27.00 27.36 28.35 27.71 Hardee 29.39 27.32 28.72 28.06 25.73 27J1 26.61 27.61 28.88 26.28 26.24 29.98 27.69 Jackson 28.57 27.53 28.04 26.58 23.28 27.68 29.08 27.90 27.93 27.52 27.39 28.15 27.64 Oadsden 27.82 27.48 26.61 26.98 23.30 28.49 28.64 28.94 29.58 26.05 26.88 27.83 27 J5 Holmes 26.93 26.99 27J7 27.66 25.97 27.72 27.94 27.61 29.87 27.19 26.02 28.43 27.49 Liberty 29.91 27.33 259S 30.81 25.79 2656 30.39 22.48 2933 25.43 29.90 24.93 27.45 Calhoun 27.64 25.83 28.11 29.18 25.43 29.18 30.17 24.18 28.93 26.00 25.^ 29.27 27.44 Jefferson 27.63 28.74 26.89 2«.07 26.01 2730 26.88 27.95 28.79 26.10 28.04 28.14 2738 Desoto 29.83 25.63 26.46 28.63 23.13 27 J6 29.89 26.52 28.13 26.54 26.00 27.89 2735 Olades 27.28 27.97 24.02 27.10 24.69 3054 28.63 27.60 28.81 26.21 27.31 27.14 2731 Okeechobee 27.13 27.10 27.46 26.47 25.13 29.09 28.55 25.80 2858 26.72 26.32 28.03 2723 Madison 27.14 26.85 26.79 26.61 26.23 28.21 28.98 25.29 29.63 28.13 25.39 27.50 2723 Bradford 26.41 27.61 26.69 28.20 26.61 27.46 29.06 26.28 27.23 25.86 27.46 26.70 27.13 Union 29.06 24.85 28.40 27.63 24.26 25.73 30.74 25.80 29.11 23.38 27.28 26.39 27.05 Taylor 24.25 28.00 23.46 27.40 29.46 26.67 27.87 25.52 27.99 27.37 25.92 27.78 26.81 Gulf 26.62 27.61 28.67 25.26 25.92 2635 27.41 23.37 24.63 28.22 30.46 25.52 26.67 Hamilton 26.64 26.88 21.76 26.02 27.87 29.47 27.03 25.58 2827 25.32 24.14 27.59 2638 Average 34.06 33.46 34.11 33.48 31.82 35.20 37.32 35.94 36.77 33.48 35.05 35.06 34.81 ^Counties are sorted by total average momentum.

PAGE 75

61 Table 3-5. Momentum index by county for years 1991 to 2002. County 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Dsde 105.28 11036 116.07 121.27 12632 131.77 137.51 143.27 149.02 154.78 160.46 165.80 Broward 104.54 109.13 113.86 118.57 12251 127.69 132.71 137.62 142.72 147.84 132.92 157.71 Orange 104.18 108.22 112.37 116.17 120.17 124.69 129.45 134.07 138.78 143.44 147.91 132.24 Palm B«ach 104.08 108.11 112.32 116.37 120.28 124.66 129.19 133.74 138.23 142.89 147.33 131.70 Hillsborough 103.98 10758 111.93 115.98 119.80 124.15 128.68 133.27 137.95 142.28 146.67 151.11 Duval 103.79 107 J2 111.48 115.19 11854 122.99 127.45 131.55 135.67 139.68 143.70 147.64 Pinellas 103.76 107.41 111.14 114.81 11830 122.16 126.28 130.23 134.18 138.23 142.13 145.92 Lee 103.40 106J6 110.04 113.47 1 16.70 120.14 123.86 127.44 131.17 134.76 138.36 142.03 Polk 103.28 106.44 109.62 112.88 115.83 119.20 122.83 126.29 129.82 133.39 136.61 140.03 Brevard 103.41 106.76 109.99 113.16 116.08 119.43 122.95 126.20 129.68 133.24 136.58 139.82 Seminole 103.15 10630 109.50 112.50 115.44 118.96 122.58 125.94 129.58 132.88 136.14 139.60 Collier 103.24 106.28 109.61 112.71 115.68 118.96 122.53 126.04 129.52 132.89 136.27 139.43 Volusia 103.30 106.40 109.59 112.68 115.67 118.93 122.38 125.58 128.99 132.36 135.58 138.80 Sarasota 103.11 106.16 109.24 112.39 115.26 118.50 122.01 123.31 128.67 131.93 135.25 138.48 Marion 103.03 106.13 109.21 112.36 115.21 118.38 121.86 125.22 128.66 131.77 134.83 138.01 Manatee 103.18 106.17 109.18 112.33 115J5 118.32 121.61 124.64 127.97 131.12 134.23 137.43 Osceola 102.95 106.07 109.14 112.10 114.90 118.03 121.22 124.62 128.02 131.07 134.08 137.32 Pasco 103.24 106.17 109.12 112.06 114.81 118.00 121.34 124.42 127.76 130.82 133.98 137.23 Lake 102.94 10555 108.98 111.96 114.80 117.91 121.27 124.45 127.82 130.88 133.95 137.03 Escambia 103.00 106.08 109.13 112.17 11454 118.13 121.44 124.54 127.77 130.81 133.78 136.88 Saint Johns 102.94 105.87 108.84 111.81 114.70 117.65 121.11 124.29 127.53 130.63 133.62 136.81 Leon 103.09 106.09 109.10 112.13 11458 118.09 121.26 124.38 127.62 130.56 133.53 136.60 Walton 103.03 10559 108.84 111.68 114 JO 117.39 120.82 123.93 127.16 130.11 133.15 136.28 Alachua 102.97 10558 108.96 111.90 114.69 117.76 120.90 123.99 127.14 130.08 133.17 136.15 Okaloosa 103.11 106.06 109.16 112.06 115.00 118.06 121.08 124.28 127.32 130.26 133.20 136.12 Santa Rosa 103.13 106.28 109.22 112.19 11453 118.04 121.23 124.29 127.18 130.12 133.10 136.03 Hernando 102.85 105.85 108.77 111.91 114.64 117.59 120.93 123.95 126.90 129.81 132.74 135.76 Flagler 102.92 105.71 108.80 111.84 114.74 117.73 120.88 123.85 127.03 129.96 132.85 135.73 C3iarlotte 103.12 10555 108.86 111.72 1 14.40 117.37 120.51 123.55 126.65 129.61 132.68 135.69 Saint Lucie 103.14 10556 108.81 111.77 114J5 117.48 120.52 123.49 126.56 129.57 132.33 133 J 8 aay 102.86 105.66 108.63 111.52 114.21 117.22 120.33 123.41 126.50 129.45 132.35 133.46 Wakulla 102.95 105.82 108.64 111.59 11431 117.51 120.84 123.80 126.87 129.72 132.68 135.46 atnis 103.14 106.01 10857 111.88 114.42 117.36 120.36 123.41 126.32 129.51 132.41 135.44 Sumter 102.88 105J9 108.35 111.08 113.79 116.85 120.14 123.26 126.36 129.46 132.39 135.30 BV 102.91 10550 108.75 111.59 11438 117.39 120.52 123.46 126.40 129.21 132.10 134.95 Martin 102.79 105J2 108.46 111.27 114.03 116.99 120.13 123.12 126.11 129.13 131.96 134.91 Nassau 102.99 105.89 108.93 111.45 114.05 116.98 120.09 122.93 125.96 129.01 131.78 134.89 Indian River 102.79 105 J6 108.25 111.21 114.01 11658 120.14 123.13 126.14 128.85 131.87 134.70 Oilchiist 103.11 105.70 108.66 111.53 114.05 117.14 120.17 123.07 125.96 128.75 131.68 134.42 Washington 103.02 105.71 108.38 111.31 11434 117.26 120.24 122.98 126.11 128.69 131.58 134.40 Columbia 102.79 105J7 108.51 111.46 1 14.20 117.20 120.31 123.14 126.04 128.83 131.55 134.36 Levy 102.85 105.71 108.56 111.43 114.00 117.00 120.02 122.89 125.82 128.58 131.45 134.28 Putnam 102.87 105.77 108.82 111.55 11435 116.99 119.82 122.61 125.63 128.37 131.24 134.08 Monroe 102.70 105J2 108.52 111.22 11351 116.73 119.68 122.67 123.38 128.15 131.07 133.85 Highlands 102.99 105.89 108.70 111.53 114.09 116.88 119.86 122.63 123.61 128.31 131.03 133.83 Hendiy 103.06 10553 108.58 111.42 114.07 116.81 119.86 122.61 125.53 128.28 130.82 133.56 Flranklin 102.94 105 JS9 108.58 111.45 11430 116.74 119.72 122.32 125.20 127.91 130.80 133.49 Suwannee 102.68 105.43 108.34 111.17 113.77 116.66 119.43 122.33 123.20 127.94 130.73 133.47 Dixie 102.46 105.42 108.42 111.26 113.67 116.43 119.25 121.94 124.79 127.68 130.59 133.46 Baker 102.77 105J2 108.14 110.92 113.47 116.23 119.27 122.07 124.95 127.81 130.59 133.43 Lafayette 102.70 10538 107.94 110.62 11339 116.24 119.08 121.86 124.98 127.68 130.41 133 J5 Hardee 102.94 103.67 108.54 111.35 11352 116.67 119.33 122.09 124.98 127.61 130.23 133.23 Jadcson 102.86 105.61 108.41 111.07 113.60 116.37 119.27 122.06 124.86 127.61 130.35 133.16 GadBden 102.78 105J3 108.19 110.89 113.42 116.27 119.13 122.03 124.98 127.39 130.28 133.06 Holmes 102.69 10539 108.15 110.92 113J1 116.28 119.08 121.84 124.83 127.55 130.15 132.99 Libetty 102.99 105.74 108.34 111.42 114.00 116.70 119.74 121.98 124.92 127.46 130.43 132.94 Calhoun 102.76 10535 108.16 111.08 113.62 116.34 119.55 121.97 124.86 127.46 130.01 132.93 Jefferson 102.77 105.64 108.33 110.94 113J4 116.27 118.96 121.75 124.63 127.24 130.04 132.86 Desoto 102.98 105J5 108.19 111.06 113J7 116.32 119.31 121.97 124.78 127.43 130.03 132.82 Olades 102.73 105J2 107.93 110.64 113.11 116.20 119.06 121.82 124.70 127.32 130.05 132.77 Okeechobee 102.71 105.42 108.17 110.82 11333 116.24 119.09 121.67 124.57 127.24 129.87 132.68 Madison 102.71 105.40 108.08 110.74 11336 116.18 119.08 121.61 124.57 127.39 129.93 132.68 Bradford 102.64 105.40 108.07 110.89 113J5 116.30 119.20 121.83 124.56 127.14 129.89 132.36 Union 102.91 10539 108.23 110.99 113.42 115.99 119.07 121.65 124.56 127.10 129.82 132.46 Ttylor 102.42 105.22 107.57 110.31 113.26 115.92 118.71 121.26 124.06 126.80 129.39 132.17 Oulf 102.66 105.42 108.29 110.82 113.41 116.04 118.78 121.12 123.58 126.41 129.45 132.00 Hamilton 102.66 10535 107.53 110.13 11252 115.86 118.57 121.13 123.95 126.48 128.90 131.66 Counties are sorted by 2002 momentum index.

PAGE 76

62

PAGE 77

CHAPTER 4 METHODOLOGY OF COMPARATIVE VALIDATION This chapter discusses all of the steps taken during our research to comparatively validate the six different forecasting techniques identified in Chapter 2. Chapter 3 outlined how momentum theory was derived and applied to strategic construction-market forecasting. This chapter applies the five remaining alternative research methods. These techniques include; 1), direct forecasting, 2) factor analysis, 3) multivariate-regression, 4) MERIC economic momentum analysis, and 5) gap analysis. A one, two and three year trend projection analysis is completed for the forecasting methods. Next, the forecasted results from the alternative methods and the momentum analysis are comparatively validated against a county's actual construction activity using simple regression. Next, all 67 Florida counties are rank ordered using their output variables and compared to their actual nonresidential permit (NRPERMIT) rank order. Finally, cluster analysis is used as a way to group the counties and is compared to two other classification techniques. There are four general steps to the comparative validation outhned in this chapter. These four steps are summarized in Figure 4-1. Figure 4-1 . Four steps for comparative validation of research methods. 63

PAGE 78

64 Alternative Research Approaches As discussed above, five alternate methods were used in our research to forecast a county's nonresidential construction activity. The following sections of this chapter describe the application and methodology of each of these approaches. Direct Forecasting The first alternative forecasting approach used to predict a county's nonresidential construction activity is direct forecasting. As discussed in Chapter 2, direct forecasting is the simplest of the forecasting methods and is the only single variable qualitative method presented in our research. Applying the technique of direct forecasting to strategic construction-market forecasting is quick and simple. The direct forecast, or estimate, is normally made fi-om a conscious or unconscious evaluation of a very small number of key demand variables that are known to most influence the required estimate, hi the case of construction-market forecasting, these estimates are based on the key indicators that would most highly correlate to a county's construction activity. There are six variables that were chosen for use with the direct forecast methodology. These six forecast variables are listed in Table 4-1. The first five variables selected are the variables from each variable group outlined in Chapter 2 that best correlated with the dependent variable nonresidential permit (NRPERMIT). These variable correlations were previously calculated during the momentum analysis in Chapter 3. The sixth variable chosen for use is the total annual value of residential permits issued in a county (RPERMIT). RPERMIT data has historically been collected and used by the industry as one of the best indicators of overall construction activity for local,

PAGE 79

65 state, and national construction-markets. The independent variable RPERMIT will be used in our research only as a variable in this direct forecast methodology. The RPERMIT variable has be excluded from the momentum analysis and all other methods in our research. Table 4-1. Six variables used in direct forecasting methodology. No. Variable Code Variable Name 1 POP Total Population 2 DVMT Daily Vehicle Miles Traveled 3 CPAYROLL Construction Payroll 4 TOTEMPLY Total Employment 5 TOTTAX Total Taxes 6 RPERMIT Residential Permits Methodology. The actual values from the six variables selected above were used as the direct forecasts, hence, a direct forecast. A rolling one, two and three year trend projection analysis was completed for each of the six variables in all of the Florida counties. This trend projection analysis methodology is described later in this chapter. The forecasted values of the six variables were compared against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6. Factor Analysis The second alternative forecasting method used to predict a counties construction activity is factor analysis. As discussed in Chapter 2, factor analysis is a generic name given to a class of multivariate statistical methods whose primary purpose is to define the underlying structure in a data matrix.' The variable data is condensed into a smaller set of factors. These factors can then be substituted for the original variables. These factors ' Hair, J. F. et al. (1998). (p. 90).

PAGE 80

66 can then be objectively compared against the results of the other forecasting methods. The statistical software SPSS Base version 1 1.5 (2002) was used to perform the factor analysis on the datasets. Methodology. The factor analysis was completed with the same ten independent variables used in the momentum analysis outlined in Chapter 2. The first step was to calculate the factor loadings for all counties. These factor loadings are linear combinations of the variables for each county and serve as an individual description of the variables. The factor loadings were then Varimax rotated and normally converged within 2 to 3 rotations. The Latent Root Criterion with an Eigen value of greater than 1 was used to determine the number of factors to extract. The set of factors explaining the greatest amount of variance in the variables was used. The computed factor scores for each county were saved as the independent variables to forecast a county's construction activity. A rolling one, two and three year trend projection analysis was completed for the computed factor scores. This trend projection analysis methodology is described later in this chapter. The forecasted values of the factor analysis were compared against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6. Multivariate-Regression The third alternative forecasting method used to predict a counties construction activity is multivariate-regression. Regression analysis is probably the most widely used and versatile dependent forecasting technique. Multivariate-regression analysis is a general statistical technique used to analyze the relationship between a single dependent

PAGE 81

67 (criterion) variable and several independent (predictor) variables.^ The objective of this method is to use known independent variables to predict a desired dependent variable. A weight for each independent variable is calculated by the regression analysis to ensure the best prediction from the set of independent variables. The weights denote the relative contribution of the independent variables to the overall prediction and facilitate interpretation as to the influence of each variable in making the prediction.^ The statistical software SPSS was used to perform the multivariate-regression on the datasets. Methodology. The multiple regression equation used for our research is shown in Equation 4-1. (Eq. 4-1) Y = |8o + |SiX, + ^2X2 . . .+ |8ioX,o The actual nonresidential permit values (NRPERMIT) were used as the dependent variable (Y). The same ten independent variables used in the momentum analysis were used for the regression independent variables (X1..10). The model coefficient /3o is the intercept and jSi. n is the model slope. The Enter regression method was used with a probability of F entry .05 and removal of . 1 0. The regression analysis was completed for all 67 counties during the years 1990 through 2002, and the model coefficients (jS's) were calculated. The resulting model was used to predict the value of NRPERMIT for all of Florida's counties during the same time period. A rolling one, two and three year trend projection analysis was completed for the forecasted values. This trend projection analysis methodology is described later in this chapter. The forecasted values of the ^ Hair, J. F. (p. 142). ^ Hair, J. F. (p. 148).

PAGE 82

68 multivariate-regression were compared against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6. MERIC Economic Momentum Analysis The fourth altemative forecasting method used to predict a counties construction activity is MERIC economic momentum analysis. As discussed in Chapter 2, the MERIC methodology measures economic momentum in a county relative to the overall economic momentum of a state. This index is a composite of percentage changes in personal income (TOTPINC), population (POP), and employment (TOTEMPLY) at the county level. An index equal to 0 means the county realized average economic growth during the decade. An index less than zero indicate relatively sluggish growth, while an index greater than zero indicates relatively prosperous growth. M etliodology. The first step of the MERIC method is to calculate the annual percentage change in the TOTPINC, POP and TOTEMPLY variables for each county using example Equation 4-2. (Eq. 4-2) Annual percent change = ((Year 2001 Year 2000) / Year 2000) This percentage change was then standardized across all counties using Equation 4-3. (Eq. 4-3) Z = X-^ a In Equation 4-3, X is the value you want to normalize, "/i" is the arithmetic mean of the distribution, and "a" is the standard deviation of the distribution. Finally, the three standardized variable estimates (i.e., TOTPINC, POP and TOTEMPLY) were then averaged for the final index score. A rolling one, two and three year trend projection analysis was completed for the forecasted scores. This trend

PAGE 83

69 projection analysis methodology is described later in this chapter. The forecasted values of the MERIC methodology were compared against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6. Gap Analysis The final alternative forecasting approach used to predict a counties construction activity is an adaptation of the Expenditure-Sales Gap Analysis that was discussed in Chapter 2 of our research. The essence of Gap Analysis is to find some way of comparing supply and demand.'* This comparison was achieved by comparing a county's construction resources (supply) with its construction activity (demand). The level of construction resources was measured by a county's total Construction Payroll (CPAYROLL) and the level of construction activity by a county's total construction permits (NRPERMIT). Methodology. The value of NRPERMIT generated for each dollar of CPAYROLL was calculated statewide and for each county by dividing NRPERMIT by CPAYROLL. Next, a county's expected payroll was calculated by dividing the NRPERMIT for each county by the average statewide NRPERMIT generated per each dollar of CPAYROLL. Finally, the gap was calculated by subtracting the actual payroll fi-om the expected payroll. An example calculation of this gap analysis is shown in Table 4-2. * Clapp, J. M. (1987). (pp. 182).

PAGE 84

70 Table 4-2. Example gap analysis calculation USDS A Total value of construction permits (NRPERMIT) Statewide $20,000,000 B Total construction employment payroll (CPAYROLL) Statewide $1,000,000 TsTRPPRMTT opnpratpH fnr parVi Hnllnr nf PPAVROT T TP = A / D Total value of construction permits (NRPERMIT) — County 1 $1,200,000 E Total construction employment payroll (CPAYROLL) County 1 $40,000 F NRPERMIT generated for each dollar of CPAYROLL (F = D / E) $30 G Expected CPAYROLL for County 1 (G = D / C) $60,000 H Actual CPAYROLL for County 1 (H = E) $40,000 I GAP (additional construction payroll needed to meet construction activity demand) $20,000 The counties were then classified by their positive gap, balanced gap, or negative gap. As discussed in Chapter 2, a positive gap indicates that additional construction resources are needed in the county to meet the construction activity demand. Balanced gap indicates a balance between the supply of construction resources and the demand of construction activity. A negative gap indicates a surplus of construction resources exist for the corresponding demand for construction activity. The three classifications of gap and their corresponding definitions are shown in Table 4-3. The findings of this gap analysis calculation and classification are included in Chapter 5. Table 4-3. Payroll gap type and definitions. Gap Definition (Gap as a percent of actual Gap Gap Type Description statewide CPAYROLL) Code Positive Demand > Supply Gap > 0.25% 2 Balanced Demand = Supply 0.25% >Gap >-0.25% 1 Negative Demand < Supply Gap < -0.25% 3 In the final step of the gap analysis, the counties were sorted in descending order based on the absolute value of the gap. This gap value was used to predict the level of construction activity in a county. A rolling one, two and three year trend projection analysis was completed for the absolute value of the gap. This trend projection analysis methodology is described later in this chapter. The forecasted values of the gap analysis

PAGE 85

71 were compared against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6. Comparative Validation of the Forecasting Methods Thus far, a total of six forecasting methodologies have been used in our research to measure and forecast a county's nonresidential construction activity. Chapter 3 detailed a new methodology of momentum analysis (including opportunity index and slope). This chapter has detailed five alternative forecasting methods including; 1), direct forecasting, 2) factor analysis, 3) multivariate-regression, 4) MERIC economic momentum analysis, and 5) gap analysis. These six methods have provided a total of twelve individual forecasts. These variables and their associated forecasts are shown in Table 4-4. The question now becomes which of these forecasting methods best predicts a county's future nonresidential construction activity (i.e., NRPERMIT). The remaining sections of this chapter will compare and statistically validate all six of these forecasting methods. Trend Analysis of Key Construction Indicators As mentioned in previous sections of our research, a rolling one, two and three year trend projection analysis was completed for the twelve output variables fi-om each of the six forecasting methods. The purpose of these trend projections is to test the various methodologies for any inherent advantages relating to the duration of the forecast. The linear trend of each output variable in a specific year was measured and projected using the method of least squares. This methodology fits a straight line to the variables over time. A minimum of five (5) known data points were used to forecast values one year, two years and three years out. The forecasted values of all methods were compared

PAGE 86

72 against a county's actual NRPERMIT values using simple regression. The standard error of the estimate and the variance were analyzed and the findings are included in Chapter 6. Table 4-4. Six forecast methodologies and their twelve associated output variables. Method Output Variable Variable Description Momentum analysis OINDEX Momentum index OSLOPE Momentum index slope Direct forecasting POP Population TOTTAX Total tax CPAYROLL Construction payroll TOTEMPLY Total employment DVMT Daily vehicle miles traveled RPERMIT Residential permits Factor analysis FACTOR Factor regression output variable Multivariate-regression LINEAR Multivariate-regression output variable MERIC economic MERIC MERIC forecast output variable momentum analysis Gap analysis GAP Gap forecast output variable Methodology for Comparative Validation The purpose of the following statistical analysis is to compare a county's actual construction activity to a county's forecasted construction activity and validate the accuracy of each method. Simple regression was used for this purpose. Simple regression is similar to the multivariate-regression method discussed earlier in this chapter. The difference is that the simple regression methodology involves the analysis of a single dependent variable (NRPERMIT) and its relationship to a single independent metric variable (forecasted activity). Simply stated, the twelve individual forecasts from the six different methodologies were each tested individually against a county's nonresidential construction activity. The simple regression equation used for our research is shown in Equation 4-4. SPSS was used to perform the simple regression analysis on the data sets. (Eq.4-4) Y = |8o + |8iX,

PAGE 87

73 The actual nonresidential permit values (NRPERMIT) were used as the dependent variable (Y). Each output variable calculated from the twelve previous forecasting methods was used for the independent variable (Xi). These variables were presented earlier in Table 4-4. The Enter regression method was used with a probabihty ofF entry .05 and removal of .10. The model coefficient /Jo is the intercept and jSi is the model slope. The regression analysis was completed for all 67 counties during the years 1996 through 2002. This methodology was used to select the forecasting method that had the highest correlation with the actual construction activity for the one year, two year, and three year trend projections. The standard error of the estimate and the statistical variance were analyzed and the findings are included in Chapter 6. During this regression analysis, the MERIC and GAP methodologies consistently demonstrated a weaker correlation to the dependent variable NRPERMIT and were not included in the following county rank variance analysis or the cluster analysis. The results of these two methodologies are included in Chapter 6 County Rank Variance The previous forecasting methods have allowed the rank ordering of the 67 Florida counties based on their output variables. These rank orders were compared to the counties actual NRPERMIT rank order. This comparison was completed for each year of the one, two, and three year trend projections and the variance was analyzed. This variance calculation measures the variance of a dataset population based on the entire population. During the preceding simple regression analysis, the MERIC and GAP methodologies consistently demonstrated a weaker correlation to the dependent variable NRPERMIT and were removed from this variance analysis. The statistical variance of

PAGE 88

74 the ten remaining method output variables was analyzed and the findings are included in Chapter 6. Cluster Analysis Overview Cluster analysis was previously selected in Chapter 2 as one of the interdependent regression techniques to be used in our research. Cluster analysis is a multivariate procedure whose purpose is to group objects based on the characteristics they possess. Cluster analysis classifies objects (i.e., Florida counties) so that each object is very similar to others in the cluster with respect to the independent variables. The resulting clusters of objects (counties) should then exhibit high internal (within-cluster) homogeneity and high external (between-cluster) heterogeneity.^ The intent of cluster analysis is the comparison of objects based on the independent variables, not an estimate of the variate. Further, the researcher is searching for the natural structure among the observations based on a multivariate profile.^ There are two primary methods of cluster analysis: (a) Hierarchical Cluster Analysis and (b) K-Means Cluster Analysis.^ The k-means method is intended to handle large research data sets with 200 or more cases, where the hierarchical method is designed for smaller data sets. Because our research includes only 67 counties, the hierarchical method was chosen. ' Hair, J. F. (p. 473). *Hair, J. F. (p. 470). 'statistical Package for the Social Sciences, Inc., Base 1 0.0 Applications Guide. (1999). Chicago, IL: SPSS Inc. (p. 293).

PAGE 89

75 Cluster analysis was used for several purposes in our research. The primary purpose was to cluster, or group, the counties by their variable characteristics and to compare these clusters to other classification techniques. Next, cluster analysis was used to evaluate the relationships between the county clusters and each individual variable (key indicator). The goal was to see if certain key indicators are better at predicting different clusters of counties. Finally, cluster analysis was used to evaluate the relationships between the county clusters and each forecasting method. The goal of this analysis was to see if certain forecasting methodologies are better at predicting different clusters of counties. Methodology for Cluster Analysis Our research has generated two primary data sets. The first variable data set includes the sixteen original variables for the 67 Florida counties over a thirteen year time period. The second methodology data set includes the twelve forecasted output variables fi-om the six different methodologies. The variable data set was used in the cluster analysis to group the counties. Due to the overall consistency in the annual trends of the research variables, and the scope limits of our research, the cluster analysis was only completed on the latest 2002 variable data set. SPSS was used to perform the cluster analysis on the variable data set. The cluster analysis was completed with the same ten independent variables used in the momentum analysis. The cases were clustered and labeled by county. A range of 2 to 8 solutions for the cluster membership was requested. Both a Dendrogram and an icicle plot were requested up to the 8* cluster range. A between-groups linkage cluster method was used with the distance measured by the Pearson's correlation interval. The measures were transformed to absolute values. The agglomeration schedule, cluster

PAGE 90

76 membership table, Dendrogram and the icicle plots were analyzed to identify the most appropriate county clusters. Construction-market classification The cluster analysis output showed that the counties could be classified into three distinct groups, or cluster memberships. Further review of the cluster analysis output and plots showed that cluster 3, the largest cluster of counties, could be broken down into three sub-clusters. This provided a total of 5 county clusters for comparison purposes. A county's cluster membership (i.e., 1 through 5) was compared to two other classification methods. These methods included gap classification (1, 2, or 3) previously outlined in this chapter, and market share classification which is outlined in the following paragraphs. A county's share of the total nonresidential construction-market was compared to a county's cluster membership and gap classification. A county's market share was calculated by dividing the total nonresidential construction activity in a specific county by the total nonresidential construction activity in the State of Florida. The counties were then classified into three market share groups. These three groups and their corresponding definitions are shown in Table 4-5. Table 4-5. County market share classifications. Market Share Level Market Share Definition (Percent of statewide NRPERMIT) Code High > 5.0% 1.0% to 5.0% < 1.0% 3 2 Medium Low The results of the gap, cluster and market share analysis were statistically compared using the Pearson correlation coefficient identified in Chapter 3. A two-tailed test of significance was used. The findings of this comparison are discussed in Chapter 5.

PAGE 91

77 Key indicators and cluster comparison Next, cluster analysis was used to evaluate the relationships between the county clusters identified above and each key indicator identified in Chapter 3. The Pearson correlation coefficient was computed between the five county clusters and the same ten key indicators used in the momentum analysis. Further, nonresidential and residential construction activity was also analyzed. A two-tailed test of significance was used. The findings of the key indicator and cluster comparison are discussed in Chapter 5. Forecasting methods and cluster comparison Each output variable from the different forecasting methodologies was also tested against the five county clusters. The simple regression equation used for this test was shown previously in Equation 4-4. SPSS was used to perform the simple regression analysis on the data set. The actual nonresidential permit values (NRPERMIT) were used as the dependent variable. As discussed earlier in this chapter, the MERIC and GAP methodologies consistently demonstrated a weaker correlation to the dependent variable NRPERMIT and were not included in this analysis. The remaining ten output variables calculated fi-om the forecasting methods were used for the independent variable. These variables were presented earlier in Table 4-4. The Enter regression method was used with a probabihty of F entry .05 and removal of .10. The regression analysis was completed for all five clusters of counties for the year 2002 only. The findings of the forecasting method and cluster comparison are discussed in Chapter 6.

PAGE 92

CHAPTER 5 KEY INDICATOR FINDINGS This chapter presents the research finding related to the key indicators of construction. The chapter begins with a review of the overall results for the key indicator analysis. This is followed by a discussion of the findings for each key indicator construct and its associated variables. Next, the results from the three construction-market classification methodologies are reviewed and classification maps are presented. Finally, the findings regarding the relationships between the key construction indicators and the county clusters are presented. In the next chapter of our research, the findings related to the various forecasting methods are presented. Overview of Key Indicators Results A total of sixteen key construction indicators were initially selected for our research in Chapter 2. The key construction indicators were grouped into six independent variable constructs. These independent variables and their associated constructs are shown in Table 5-1. Table 5-1. Key construction indicator constructs and associated variables. Population Geographic Initial Employment Economic Financial Advantage Infrastructure Transition Environment Resources POP PROX DVMT CPAYROLL TOTEMPLY TOTTAX PDENSITY CLMILES TOTPINC GSALES ALVALUE RDENSITY AVEWAGE PLINDEX TOTREV RPERMIT In Chapter 3, the Pearson correlation was computed for all of the independent variables for the years 1990 through 2002. The results of this analysis are summarized in Table 5-2. The indicators have been sorted by their average correlation to NRPERMIT 78

PAGE 93

79 and include the variable description and group. Eleven of the 16 variables were highly correlated ( > 0.800) to the dependent variable nonresidential permit (NRPERMIT). These eleven variables have Pearson's Coefficient averages ranging fi-om 0.829 to 0.962. The remaining five variables were then dropped fi-om use in the research due to their low correlation (< 0.800) to NRPERMIT. These five excluded variables are shown below the dashed line in Table 5-2. In addition to these five excluded variables, the independent variable RPERMIT was used in our research only as a variable in the direct forecast methodology. The RPERMIT variable was excluded fi-om the momentum analysis and all other methods in our research. Table 5-2. Pearson correlation results for each research variable. Variable Average Variable Group Correlation NRPERMIT 1.000 TOTTAX 0.962 Financial Resources ALVALUE 0.960 Financial Resources TOTEMPLY 0.948 Economic Environment GSALES 0.941 Economic Environment CPAYROLL 0.940 Employment Transition DVMT 0.932 Initial Infi-astructure POP 0.924 Population TOTPINC 0.912 Employment Transition TOTREV 0.856 Financial Resources RPERMIT 0.854 Initial Infi-astructure CLMILES 0.829 Initial Infrastructure AVEWAGE 0.660 Employment Transition PLINDEX 0.637 Economic Environment PDENSITY 0.624 Population RDENSITY 0.458 Initial Infrastructure PROX 0.360 Geographic Advantage The following sections of this chapter analyze the key variable correlations to NRPERMIT by using the six independent variable constructs previously discussed. Remember that these constructs are nothing more than logical groupings of the key

PAGE 94

80 construction activity indicators identified in the preceding literature review. As discussed in Chapter 2, using these constructs as descriptors of the relationship between the key indicators and a county's construction activity allows for a more understandable and vivid discussion. Financial Resources Overall, the financial resources group of key indicators was the most predictive of nonresidential construction activity. As shown in Table 5-2, two of the three financial resource variables, (TOTTAX and AL VALUE) had the highest correlation with NRPERMIT and the third finished in the top eleven. This is a fairly logical finding because the financial resources must be available to the marketplace to construct new facilities and infi-astructure. As was suggested in Chapter 2, a market's access to financial resources is positively correlated with a county's construction activity. Economic Environment The second most predictive group of key indicators appears to be related to a county's economic environment. As shown in Table 5-2, two of the five highest correlating variables were in this group. A county's economic environment must be conducive to the growth of investment and employment. The improvement of the economic environment within a county was found to be positively correlated with a county's construction activity. One dissenting variable from this group was a county's price level index (PLINDEX). The price level index is a set of numbers which reflects the price level in each county relative to populationweighted statewide average (100 for each category) for a particular point in time. It measures price level differences fi-om place to place in contrast to the consumer price index prepared by the U.S. Bureau of Labor Statistics,

PAGE 95

81 which measures price level changes from month to month.' The basket of goods measured in the price level index includes housing, food and beverage, health care, transportation and other miscellaneous goods and services. It appears that this variable is not applicable at the county level because it uniformly affects all of the counties in the state, or the effect of "a rising tide raises all boats." Employment Transition The next most predictive group of key indicators appears to be employment transition. This group of variables is closely related to the economic environment construct discussed above and includes the county's transition into certain types of higher paying employment. As discussed in Chapter 2, shifts from the lower income employment sectors to the higher income sectors will stimulate investment in new facilities and infrastructure. This employment transition from low to high-income employment sectors was found to be positively correlated with a county's construction activity. One dissenting variable from this group was a county's average wage (AVEWAGE). Average wage is the average earnings per job in dollars. There is no clear reason for the lower correlation particularly when average wage was highly correlated with the best predictor, a county's total tax revenue (TOTTAX). Average wage seemed to predict best in the most populated counties. Initial Infrastructure The next most predictive group of key construction indicators appears to be related to a county's initial infrastructure use and investment. Table 5-2 shows two of the three ' Bureau of Economic and Business Research. (2002). Florida Statistical Abstract 2002. Gainesville: University of Florida, Warrington College of Business Administration, (p. 761).

PAGE 96

82 variables correlating highly with NRPERMIT. As discussed in Chapter 2, public construction projects tend to become larger and more complex as a county develops. Population growth drives the demand for residential housing, water resources, and energy delivery systems. Unpaved roads are paved, and existing road capacity is increased. This initial construction activity is a predecessor to larger public construction projects such as toll highways, power plants, and water and wastewater treatment facilities. The increased use and growth of a county's roadway infrastructure was found to be positively correlated with a county's nonresidential construction activity. One dissenting variable from this group was a county's road density (RDENSITY). Road density was computed by dividing a county's total centerline miles of roadway by its total land area (miles/square mile). The reason for this low correlation to NRPERMIT may be that a county's land area is simply the area within a geographic boundary and that this measure is independent and unrelated to either roadway miles or construction activity. Population Finally, and most unexpectedly, a county's total population (POP) was found to be only an average predictor of nonresidential construction activity in comparison to the other key variables. It was hypothesized that increases in a county's total population should drive the need for new buildings and infrastructure. But why was the population variable correlation so low? One reason is that this variable may really be a function of the greater economic opportimity available in a geographic region, not the driver. A healthy economic environment generally creates more employment opportunity. This environment increases the demand for more and better skilled workers which in-tum drives higher wages. This opportunity and higher

PAGE 97

83 income potential attracts population to the area. This was evidenced by the population (POP) variable's very high correlation to a county's total employment (0.981) and total personal income (0.979), but its lower correlation to NRPERMIT (0.924). Further, a county's population density (PDENSITY) was not found to be a good predictor of construction activity for the same reasons previously discussed for road density. Geographic Advantage A county's proximity to major urban areas (PROX), or geographic advantage, was not found to be a good predictor of nonresidential construction activity. In Chapter 2, it was discussed how secondary urban areas, or bedroom communities, typically develop around larger cities. It was anticipated that this spillover effect would contribute to the development of adjacent counties and would be further amplified if the county is located between multiple large cities. To test this hypothesis, a proximity factor was constructed that measured the relationship of proximity and population between counties (see Chapter 2 for details). This factor provided a value that is weighted both by distance and population. While this factor did not predict well overall, it did perform well on the most populated counties, counties immediately adjacent to large cities, counties between larger counties, and counties adjacent to non-coastal urban centers like Orlando. There are several suggestions for improving this factor. First, using more than the top 10 most populated counties would allow counties such as Escambia (Pensacola) and Leon (Tallahassee) to be included in the calculation. This may help improve the results of the counties located throughout Florida's panhandle because the most populated counties are generally found in South Florida. Second, the calculation could be modified to include more than the 2 closest and most populated counties. This may help improve the results

PAGE 98

84 of the counties that are not located directly between two large counties. Finally, the distance used to calculate the factor was the distance between the county seats. It may be more appropriate to use the distance between the centroids of a county's land area or population. Construction-Market Classification Three different methodologies were presented in Chapter 4 that can be used to classify counties. These methods included cluster analysis, gap analysis and a county's percent share of the total nonresidential market. The classification results of these three methods are shown in Table 5-3 for all 67 Florida counties. To provide a geographical context for these findings, the results from Table 5-3 were mapped and are shown in Figure 5-1. The correlation matrix between these three methods is shown in Table 5-4. The classification method inter-correlations ranged fi-om 0.647 to 0.762. While these correlations may appear relatively low, it should be understood that these different classification methods are fundamentally different measures by the nature of their variables. Viewing the counties through these three classification methods has advantages over using just one classification method. The variables and advantages of each method are discussed in the following paragraphs.

PAGE 99

85 Table 5-3. Results of the cluster, gap, and market share classification methods. Cluster Gap Percent Cluster Gap Percent Analysis Type Market Analysis Type Market County^ Membership Code Share County Membership Code Share Duval 5 3 3 Hardee 2 1 1 Lee 5 3 2 Hendry 2 1 1 Dade 5 2 3 Indian River 2 1 1 Broward 4 3 3 Nassau 2 1 1 Hillsborough 4 3 3 Okeechobee 2 1 1 Orange 4 3 3 Taylor 2 1 1 Collier 4 3 2 Flagler 1 2 1 Pinellas 4 3 2 Santa Rosa 1 2 1 Seminole 4 3 2 Sumter 1 2 1 Palm Beach 4 2 3 Baker 1 1 1 Manatee 4 2 2 Bradford 1 1 1 Sarasota 4 1 2 Calhoun 1 1 1 Monroe 4 1 1 Columbia 1 1 1 Brevard 3 3 2 Dixie 1 1 1 Escambia 3 3 2 Franklin 1 1 1 Lake 3 3 1 Gadsden 1 1 1 Alachua 3 2 2 Gilchrist 1 1 1 Leon 3 2 2 Glades 1 11 Marion 3 2 2 Hamilton 1 1 1 Pasco 3 2 2 Hernando 1 11 Saint Lucie 3 2 2 Holmes 1 1 1 Polk 3 1 2 Jackson 1 1 1 Volusia 3 1 2 Jefferson 1 11 Bay 3 1 1 Lafayette 1 1 1 Citrus 3 1 Levy Clay 3 1 Liberty Desoto 3 1 Madison Highlands 3 1 Putnam Okaloosa 3 1 Suwannee Saint Johns 3 1 Union Osceola 2 2 Wakulla Martin 2 2 Walton Charlotte 2 1 Washington Gulf 2 1 1 Counties are sorted by clustei •, then gap, then percent market share results.

PAGE 100

86 U Ah O u o o o o o o ^ 13 o ^ o VO O o ^ o c/1 13 o o o o o 1/1 13 o U 00 u ^ I c o IT) 0) 1 (/} o Oh '3^ 3 (A -4-* CO 3 CO B c , 1? CO >> O '3 J 13 d :3 o o o (» 9>

PAGE 101

87 The cluster analysis method classified the counties based on the characteristics of ten independent variables. By nature of the statistical model, the resulting clusters of counties should then exhibit high internal (within-cluster) homogeneity and high external (between-cluster) heterogeneity. This method of classification would be useful to large design and construction (D&C) firms trying to target counties with similar attributes to counties where they have had previous business success. In contrast, the gap classification method used only two supply and demand variables which estimate the gradient, or topography, of construction payroll gap over a geographic area (Florida counties). This methodology would be useful to a D&C firm that is evaluating areas of supply and demand imbalance. An example of how this is used is if a D&C firm was trying to decide whether to open a new office in either Broward or Dade County. Table 5-3 and Figure 5-1 show that both counties have greater than 5% of the total state nonresidential construction-market share, but the ratio of demand to construction resource supply is more attractive in Dade County. The third method, percent market share, could be considered a direct classification method. This method classifies the counties into three groups by simply using three levels of the dependent variable NRPERMIT as a percentage of the state total. This method would most likely be used in conjunction with one of the two previous methods and would act as a gauge of the overall magnitude of construction activity in a county. An example of how this is used is if a D&C firm was trying to decide whether to open a new office in either Hillsborough or Pinellas County. Table 5-3 and Figure 5-1 show that while both counties appear equal in the cluster and gap classification methodologies,

PAGE 102

88 Hillsborough is the better choice due to the higher percent market share (i.e., construction activity). Key Indicators and County Clusters In Chapter 4, cluster analysis was described as a way to evaluate the relationships between the clusters of counties and each key indicator. The Pearson correlation coefficient was computed between the five county clusters and the same ten key indicators used in the momentum analysis. NRPERMIT and RPERMIT were also included. The average value of each key variable within each cluster was also calculated. Due to the overall consistency in the annual trends of the independent variables, and the scope limits of our research, the cluster analysis was only completed on the latest 2002 variable data set. The findings of the key indicator and cluster comparison are provided in Table 5-5. Table 5-5 lists the key indicators down the columns, and the clusters across the rows. The counties included in each of the five clusters are shown along with the corresponding Pearsons coefficient. The average value for each key construction indicator in each cluster is also shown. The overall average correlation of the key indicators was found to be relatively high in clusters 2 through 5 with coefficients ranging fi-om a high of 0.992 to a low of 0.849. All key indicators correlated below average (0.710) in county cluster 1. Residential permit activity (RPERMIT) produced the lowest correlation across all of the clusters with an average Pearson's coefficient of 0.631. The second lowest correlating variable across the county clusters was centerline miles of roadway (CLMILES) with a coefficient of 0.703. All other key indicator coefficients across the clusters were above 0.861.

PAGE 103

89 The average key indicator values consistently decreased down the clusters with the highest average values found in cluster 5 and the lowest values in cluster 1 . The lowest average indicator values and the lowest correlations were both found in county cluster 1 . These findings are further validated by the variance analysis results presented in Chapter 6. It appears that the counties with the lowest key indicator values are the least predictable in regards to nonresidential construction activity.

PAGE 104

90 O O .2 c o o o 1/1 T3 CO O • t-H C o 3 t/J u Pi in s ° C; 5 -a 6 > g Q «) M q O (5 w w o — — a i « (3 a 2 2 Eg O p) (% vj V) > s 2 a _ J, » g ss » 5 PQ (5 o G Q (3 ; -a 2 P o pS S) J ^ ^ 21 O O H I 6 8 -9 « a a '4 00 * 3 a a > (vj <7\ o C3 C3 O >o 0\ •— S § £ o — o ^ s o o o tn <^ " «|-| 00 >o ' 00 o <^ b~ o\ oa o ci o o o o 7 S; S r
PAGE 105

CHAPTER 6 FORECASTING METHODOLOGY FINDINGS This chapter presents the research finding related to the comparative validation of the various forecasting methods presented in Chapters 3 and 4 of our research. The chapter begins with a review of the overall results for the six forecast methods including a discussion of the R^, standard error of the estimate, F, and t statistics. This is followed by a review of the finding for each individual forecast methodology. Four of the six forecasting methods tested were found to be very good predictors of nonresidential construction activity. The forecasts for the best nonresidential construction-markets in the State of Florida for the year 2005 are presented. Next, the statistical variance results between the projected county rankings and the actual county rankings are presented. Finally, the findings of the comparison between the forecasting methodologies and the county clusters are reviewed. Statistical Overview Six forecasting methodologies were comparatively validated in Chapter 4. These six methodologies included; 1) momentum analysis, 2) direct forecasting, 3) factor analysis, 4) multivariate-regression, 5) MERIC economic momentum analysis and 6) gap analysis. These forecasting methodologies generated twelve output variables (Table 4-4). The twelve output variables were analyzed against a county's actual nonresidential permit activity (NRPERMIT) using simple linear regression analysis. The same analysis was completed for each of the 1, 2, and 3 year trend forecasts. Details of the regression analysis methodology were provided in Chapter 4. A summary of the results of this 91

PAGE 106

92 analysis are presented in Table 6-1. Table 6-1 has been sorted by the average standard error of the estimate and includes the average The average values for the forecast methodology are .796, .790, and .783 for the one, two, and three year trend projections respectively. These values indicate that overall the methodologies explain a large proportion of the total variation of NRPERMIT, and indicate a strong linear relationship between the dependent and independent variables. The decrease in the average R^ values indicates a weaker relationship as the length of the trend projections increase. Adjusted R^ was not significantly different fi-om R^ because there was only one predictor selected for each regression. The average values for the standard error of the estimate are approximately $169, $179, and $191 million for the 1, 2, and 3 year trend projection respectively. The standard error of the estimate was compared with the standard deviation of NRPERMIT and was found to be consistently less. This indicates that the regression models overall are better than the mean as a predictor of the independent variable NRPERMIT. One, Two, and Three Year Trend Projections For the 1 year projection of NRPERMIT shown in Table 6-1, nine of the twelve forecast method output variables have average R^ results ranging from 0.871 to 0.942 with an average standard error of the estimate ranging fi-om approximately $102 million to $150 million. The remaining three output variables (shown below the dashed line) were significantly less accurate with the highest R^ reaching only 0.724 and with a standard error more than two to four times the lowest standard error.

PAGE 107

93 0 tS n o 'I •8 > t O 0 w-1 0\ On •ri 0 cs 0 0 0 ON NO 0 o\ ON o\ On On On 00 00 d d d d d d d d d X w 3 W Q; 2 Q o S M |j o o o Q a, 2 O 2 — 00 vo On «t^ ^« "T^ On" fO On <^ C-« M ^ vi t-1/1 t-' ^ NO m 1U-l NO o^ On^ 00^ ^ O ro — . 00 On o 10 O ir op U a, 2 < w 5 S S S < no «-l T) WN NO ^ 00" — " 00' ^0*0 o o 6

PAGE 108

94 For the 2 year projection of NRPERMIT, the same top nine output variables have an average ranging from 0.864 to 0.945 with an average standard error of the estimate ranging from approximately $104 million to $159 million. As with the 1 year projection, the remaining three output variables were significantly less accurate with the highest R reaching only 0.745 and with a standard error again more than two to four times the lowest standard error. As expected, there is an overall decrease in the average R values and an overall increase in the standard error which indicates a weaker relationship as the length of the trend projections increase. , 2 For the 3 year projection, the same top nine variables had a similar average R and standard error range. As with the 1 and 2 year projections, the remaining three output variables were significantly less accurate and demonstrated the same overall decrease in average R values and increase in the standard error. This again confirms a weaker predictive relationship as the length of the trend projections increase. Validation of Statistical Signiflcance The F and t statistics were analyzed and the results are shown in Table 6-2. The methods were sorted by their average F statistic. The F statistic was used to test the hypothesis of the slope. The average values for F are 628, 604, and 560 for the one, two, and three year trend projections respectively. These large F values indicate that overall the methodology variables sfrongly explain the variation of the independent variable. The decrease in the average F values indicates a weaker relationship as the length of the trend projections increase. TOTTAX provided the highest F with an average value of 1,164, 1,156, and 1,073 for the one, two, and three year frend projections respectively. The three methodologies with the lowest F are RPERMIT, GAP, and MEiaC. The average F statistic for MERIC is only 2 which

PAGE 109

1 95 indicated a very weak explanation of the independent variable variation. These results are consistent with the and the standard error of the estimate results previously discussed. The / statistic was used to test the significance of the slope. The average values for t are 22.799, 22.307, and 21 .549 for the one, two, and three year trend projections respectively. These large t values indicate the correlation between the dependent and independent variables are statistically significant, and are not caused by sampling variability. The slight decrease in the average / values indicates a weaker relationship as the length of the trend projections increase. TOTTAX provided the highest t statistic with average values of approximately 33.857, 34.167, and 32.800 for the one, two, and three year trend projections respectively. The three methodologies with the lowest t are RPERMIT, GAP, and MERIC. The average t statistic for MERIC is only 1 which indicates the correlation between the dependent and independent variables are not statistically significant. These results are also consistent with the and the standard error of the estimate results previously discussed. To further check the validation of the results, a split sample regression was preformed on the output variables from several of the forecast methodologies using the same simple linear regression analysis. There were no significant differences found during this validation check.

PAGE 110

96 I 00 I I II 5 ooooooooo ooooooooo o 5 o EJ H c <; c/3 O £ O J o ^ <: 0. Q U 2 O 00 M o >n m 00 ON 00 3 o o ON p-j r Gc — OO\00t-t-NOvO>n o< 00 •S 3u a 2 6 o X pa w a, Q O S « o o o g u 2

PAGE 111

97 Accuracy of Forecasting Methodologies The following sections of this chapter present the findings of the comparative validation for each of the six forecasting methodologies presented in Chapters 3 and 4. Table 6-3 summarizes the relationship between each of the twelve trend forecasts and the average standard error, technique classification, and the variable relationship. Direct Forecasting The methodology of direct forecasting was found to be the best predictor of nonresidential construction activity in our research. As outlined in Table 6-3, direct forecasts consistently provided the best and fourth best forecasts for the 1, 2, and 3 year trend projections. As discussed in Chapter 2, direct forecasts are the simplest of the forecasting methods. The direct forecast variables in our research estimated the value of NRPERMIT with a single variable and without any intervening steps. Direct forecasts are gut feelings based on industry conventional wisdom or learned experience. As shown in Table 6-3 and previously discussed in Chapter 5, a county's total tax revenue (TOTTAX) and total employment (TOTEMPLY) were the best direct indicators respectively. The first, and probably biggest surprise of our research, was the direct forecast using a county's residential permit activity (RPERMIT). RPERMIT was the least accurate of the six direct forecasts and provided only the 10* best overall forecast for the 1, 2, and 3 year trend projections. RPERMIT is the total value of a county's new residential buildings authorized by building permits. Starting in 1996, the 67 counties in the State of Florida were no longer required to report on total and nonresidential building permit data. It was thought that residential permit data was an accurate reflection of the overall construction activity within a county. While RPERMIT does correlate well to

PAGE 112

98 total building permit activity (R^ approximately 0.925), RPERMIT has a much weaker correlation to nonresidential building permit activity (R^ approximately 0.729). This is concerning because nonresidential building permits accounted for over 45% of the total construction permits authorized in the State of Florida from 1990 through 2002. Our research has found that residential permit data is one of the least accurate ways of predicting nonresidential building activity in the State of Florida relative to the other methods tested. Factor Analysis The second most accurate type of forecasting method appears to be the statistical method of factor analysis. Factor analysis finished 2"** on average for the 1, 2, and 3 year trend projections. Factor analysis was the second most complex forecasting methodology used in our research next to multivariate-regression. Factor analysis is one of the three interdependent forecasting techniques used in our research, and was found to be the most accurate. An interdependent technique is one in which the prediction equation is not dependent on, or estimated using, known dependent variable values. It involves the simultaneous analysis of all the variables in the set. Based on the accuracy of its prediction, this methodology did a very good job of identifying the underlying structure of the highly correlated data used in our research. But again, why is a simple direct forecast approach like TOTTAX more accurate than this more complex method? The most likely reason is that a direct methodology is not burdened by lower correlating variables that are inherent in a factor regression model.

PAGE 113

99 o S a) W •s g M T3 cal cal Direct Direct O Q Direct Direct Stati! Stati! Detei Detei Dire( CO C :5 -a <^ -3 " C Oh T3 •(75 ^ Q *J 4_. « ll> U c c 2 3 3 " CO ia ia a, a, i> l> r> VI m til St U !-> 4-1 U bp OO f<1 VO o m m CI 0\_^ \0 o) oo On •O vo" 'O O o r-* oo" (N t-~" o «r> Q PO NO lO 3: CO riinvo-^r~No>o u-i«r~ooN(NOiot^ No_^ f^. "T, '^i, ^ ^ 7-" i/t' m" r~" no" no" r--" in" oC oo oo NO O 1/1 f«1_ 0n_ (N m" Tf" o" oo Tf N

PAGE 114

100 Multivariate-Regression The next most accurate type of forecasting method appears to be multivariateregression. Table 6-3 shows the multivariate-regression methodology provided the S"^** most accurate forecast on average for the 1, 2, and 3 year trend projections. As discussed in Chapter 2, multivariate statistical regression is the most complex of the forecasting methodologies used in our research. Multivariate-regression is one of three forecasting methodologies in our research that can be classified as a dependent technique. This means that the multivariate-regression prediction equation was dependent on, and estimated using, known dependent variable values (NRPERMIT). While this may explain the higher overall accuracy of this forecasting method relative to the other methods, it is unclear why a simple direct forecasting method like TOTTAX is more accurate than this complex method. The most likely reason is the same as previously given for factor analysis in that a direct methodology is not burdened by the lower correlating variables that are included in the multivariate-regression model. Momentum Analysis The next most accurate forecasting method appears to be the new momentum analysis index and its corresponding slope. As shown in Table 6-3, both momentum analysis forecasts consistently finished 5* and 6* for the 1, 2, and 3 year trend projections. Momentum analysis is a dependent approach that was designed to be less complex but more meaningfiil than the traditional statistical regression approaches. Traditional regression analysis equations are constructed with a linear combination of one-dimensional variables with empirically determined weights. The momentum analysis equation is also constructed with a linear combination of variables but the difference is that each variable includes three dimensions. These dimensions include the

PAGE 115

101 relative size, rate of change, and influence of the variable. These three dimensions are combined into one new variable value called momentum. Based on the accuracy of its prediction (R^ ranged from 0.897 to 0.909), this momentum methodology clearly did a good job of forecasting a county's construction activity. But then again, why is a very simple and direct forecasting approach more accurate than this more complex method? There may be a couple reasons. The first is the same as previously outlined for factor and multivariate analysis regarding the burden of lower correlating variables. A second reason relates to the added variable dimensions. While this multidimensional approach is one of the strengths of momentum analysis, more research needs to be done regarding the appropriate weights of each dimension of the variable. As will be seen with the MERIC forecasting method discussed below, velocity, or the rate of change of the variable can be detrimental to the accuracy of the forecasting model if incorrectly measured or weighted. Preliminary experimentation during our research showed an improvement in the momentum prediction if the velocity dimension was mathematically suppressed using a weight or coefficient. Conversely, if the size (mass) dimension of the variable is enhanced, there is a corresponding improvement in accuracy of the momentum forecast. Second, the calculation of the momentum index itself may be affecting the accuracy of the forecast. The momentum index is a measure of the cumulative value of a county's total momentum. As a county's armual momentum is calculated, it is added to the previous year's momentum index. The prediction error of the new momentum forecast could possibly be compounded by accumulating error from the previous forecasts.

PAGE 116

102 Evidence of this is how the forecasting accuracy of momentum analysis may have declined slightly throughout the years of the forecast. Gap Analysis One of the most inaccurate forecasting methods appears to be gap analysis. Gap analysis consistently finished with the second worst prediction for the 1, 2, and 3 year trend projections. Gap analysis is a two variable interdependent approach that compares a county's construction resources (supply) with its construction activity (demand). CPAYROLL was used as the supply variable and NRPERMIT was used as the demand variable. Why was gap analysis so inaccurate in forecasting NRPERMIT? The most likely reason is that the output variable fi-om the gap analysis equation is simply a ratio between two independent variables. The gap calculation itself eliminates any effect of the variable's initial size and subsequently its scale to NRPERMIT. While gap analysis did not perform well in predicting the level of construction activity in a county, our research used gap analysis for two other purposes. As outlined in Chapter 5, gap analysis was used to identify the Florida counties with construction demand that is larger or growing more rapidly, than the available supply. Second, gap analysis was used as a tool to classify the 67 Florida counties. Results of these two applications of gap analysis were discussed in the previous chapter. Missouri Econoinic Research and Information Center Economic Momentum Analysis Finally, the most inaccurate forecasting method was the MERIC economic momentum analysis. Table 6-3 shows MERIC 's (Missouri Economic Research and Information Center's) momentum methodology consistently finished with the worst prediction for the 1, 2, and 3 year trend projections. This MERIC index is simply an

PAGE 117

103 average of normalized percentage changes in personal income, population, and employment in a county. So, why was this method so inaccurate in forecasting NRPERMIT? There may be two primary reasons. First, the MERIC index variables were intended to be a measure of economic momentum in a county relative to the overall economic momentum of a state. The MERIC index was not intended to measure a specific sector of the economy (i.e., construction) within a county. Second, as discussed previously in this chapter, the rate of change of a variable can be detrimental to the accuracy of the forecasting model if incorrectly measured. The rate of change in the original MERIC model was measured in 5 year periods. Due to the design of our research, the rate of change was required to be measured every year. This shorter frequency of time generated a more volatile rate of change whereas a longer frequency of time between the data points creates a bridging effect. This effect acts to minimize sudden variations in the subsequent forecast. Further, this frequency of measurement affects each variable differently. A relatively steady variable such as population would not be impacted as much as the more volatile income and employment variables. Where are the Best Construction-Markets? One of the fundamental questions of our research was to identify the location of the best new construction-markets. Based on the findings of our research, a direct forecast using a county's total tax revenue (TOTTAX) was consistently found to be the best predictor of nonresidential construction activity in the State of Florida. As discussed in Chapter 4, a three year trend projection analysis was completed for the variable TOTTAX. The linear trend of TOTTAX was measured and projected using the method of least squares. Five known data points were used to forecast the TOTTAX values three years out. This trend projection resulted in a rank ordering of the 67 Florida

PAGE 118

104 counties and the findings are shown in Table 6-4. To provide a geographical context for these findings, the 2005 three year trend projection results fi^om Table 6-4 were mapped and are shown in Figure 6-1. As shown in Table 6-4, the top five nonresidential construction-markets in the State of Florida for the years 2004 and 2005 are forecasted to be; 1) Dade, 2) Broward, 3) Orange, 4) Palm Beach, and 5) Hillsborough counties. The top 15 counties were very consistent in their rank throughout the study period with an average statistical variance of only 0.5 (0.0 = no rank variance). The counties ranked 16'*' through 40"^ were also consistent in their rankings with some exceptions and have an average statistical variance of 3. 1 . The remainder of the counties demonstrated an overall increase in rank fluctuation with an average statistical variance of 7.9. One general observation is the apparent increase in rank fluctuation for all the counties during in the years 2003 through 2005. Depending on a county's sources for tax revenue, this could be attributed to the economic recession that began three years earlier in 2000. County Rank Variance The forecasting methods presented in our research have allowed the rank ordering of the 67 Florida counties based on their output variables. These rank orders were compared to the counties actual NRPERMIT rank order. The statistical variance was analyzed for each of the 1, 2, and 3 year trend projections. The total variance for each year is presented in Table 6-5. As expected, there is an overall increase in the total variance for the 1, 2, and 3 years trend projections. This confirms that the overall variance in the forecasts increase

PAGE 119

105 as the length of the trend projection increases. Also, the variance for year 2002 was found to be consistently less that of the preceding years. Figure 6-1. Map of the 2005 total tax-direct forecast results using a three year trend projection. Further analysis of the county rank variance findings show that as a county's nonresidential construction permit activity increases, the variance in a county's forecasted rank decreases. This is evidence that a county with a larger market share is more predictable, and a county with a smaller market share is less predictable. This relationship between rank variance and construction-market share is apparent in the example graph of the 2002 three year trend variance analysis shown in Figure 6-2.

PAGE 120

106 Table 6-4. Rank of nonresidential construction-markets as forecasted by total tax collections using a 3 year trend County' 1997 1998 1999 2000 2001 2002 2003 2004 2005 Dade 1 1 1 1 1 1 1 1 1 Broward 2 2 2 2 3 3 3 3 2 Orange 3 3 3 3 2 2 2 2 3 Palm Beach 4 4 4 4 4 4 4 4 4 Hillsborough 5 S 5 S 5 S S S S Duval 7 7 6 6 6 6 6 6 6 Pinellas 6 6 7 7 7 7 7 7 7 Lee 8 8 8 8 8 8 8 a s A o Volusia 13 12 12 11 12 13 13 14 9 Brevard 10 10 11 12 13 11 11 10 10 Polk 9 9 9 9 10 9 9 9 11 Collier IS 14 14 14 14 14 14 13 12 Sarasota 12 13 13 13 11 12 12 12 13 Soninole 11 11 10 10 9 10 10 11 14 Manatee 16 16 17 17 17 16 18 17 IS Escambia 14 IS 15 15 15 IS IS 16 16 Pasco 18 19 19 19 19 19 19 19 17 Marion 20 18 18 18 18 17 17 18 18 Leon 17 17 16 16 16 18 20 20 19 Osceola 23 23 23 21 20 20 22 23 20 Alachua 19 20 20 20 21 21 23 22 21 Okaloosa 21 21 21 39 31 24 21 IS 22 Martin 25 2S 24 25 2S 26 16 21 23 Lake 26 26 26 24 24 25 24 24 24 Bay 22 22 22 22 22 22 2S 25 2S Monroe 24 24 25 23 23 23 26 26 26 Saint Lucie 27 27 27 26 27 28 30 27 27 Charlotte 28 28 28 27 29 30 27 28 28 Saint Johns 30 29 29 29 28 29 28 29 29 Indian River 31 31 31 28 26 27 29 30 30 Clay 29 30 30 30 30 31 31 31 31 Citrus 32 33 33 32 33 33 33 33 32 Hernando 33 32 32 31 32 32 34 34 33 Highlands 34 34 3S 33 36 35 32 32 34 Santa Rosa 35 35 34 34 34 34 36 36 35 Walton 38 38 37 36 35 36 35 35 36 Columbia 36 36 36 35 37 37 38 38 37 Nassau 37 37 39 37 38 38 37 37 38 Flagler 41 41 41 41 42 40 42 40 39 Putnam 39 39 38 38 39 39 39 39 40 Hendry 45 43 43 43 44 44 43 41 41 Sumter 43 45 45 45 4S 46 46 44 42 Okeechobee 42 42 42 42 43 42 44 42 43 Jackson 40 40 40 40 40 41 45 43 44 Suwannee 44 46 46 47 47 47 48 48 45 Levy 47 47 47 44 46 45 47 47 46 Gadsden 46 44 44 46 48 48 50 49 47 Bradford 51 51 50 50 50 49 52 51 48 Madison 57 58 60 59 58 57 41 45 49 Desoto 48 49 49 48 41 43 49 50 50 Hardee 49 50 51 51 51 51 53 52 51 Taylor 50 48 48 49 49 50 51 53 52 Baker 55 55 55 54 S3 52 56 54 S3 Franklin 54 S3 53 56 56 56 58 55 54 Dixie 62 62 62 62 64 63 40 46 55 Washington 52 52 52 53 54 53 57 56 56 Jefferson 53 54 57 61 62 59 60 59 57 Wakulla 56 57 58 57 55 54 59 57 58 Holmes 59 56 59 60 59 58 54 58 59 Gulf 58 60 56 55 57 61 64 61 60 Gilchrist 63 63 63 64 63 64 65 62 61 Calhoun 61 61 61 58 60 60 62 63 62 Lafayette 67 67 67 67 67 67 55 60 63 Union 64 64 64 63 61 62 63 64 64 Glades 66 65 66 66 66 66 66 66 65 Liberty 65 66 65 65 65 65 67 67 66 Hamilton 60 59 54 52 52 55 61 65 67 " Counties are sorted by their 2005 rank.

PAGE 121

107 Table 6-5. Total county rank variance for forecast years 1996 through 2002. Forecast Year lY 2Y 3Y Average 1996 1,641 1,641 1997 1,774 1,781 1,778 1998 1,678 1,729 1,725 1,711 1999 1,654 1,807 1,835 1,765 2000 1,705 1,704 1,795 1,735 2001 1,533 1,723 1,728 1,661 2002 1,377 1,383 1,397 1,386 Average 1,623 1,688 1,696 /A BNi k < J \ S /M\ 1 ni|ifiMMiMliilii|iiMMt||iUiMiHii|nii|iii|iiH||ii^ I" I' i I" * 1 o S loS I 'I'll".! =3IJ;S Oil i|o '||j |3|-' Countias Sorted by Total Valuo of Permits Figure 6-2. County rank variance comparison for the year 2002. In this figure, the 67 Florida counties are shown along the X axis and are ranked by their actual nonresidential construction activity in descending order. This creates a straight line graph that begins at rank 1 in the lower left of the graph and rises to rank value 67 in the upper right. The forecasted ranks calculated from the methodologies previously discussed in this chapter were overlaid onto the same graph. It can be seen that the forecasted rankings wrap tightly around the counties with the highest actual construction activity (e.g., Dade Palm Beach, Broward) and exhibit little variance. Conversely, the forecasted rankings exhibit an increase in variance as they move to the right with the counties that have the lowest construction activity.

PAGE 122

108 A closer analysis of Figure 6-2 shows that several counties (e.g., Saint Lucie, Sumter, and Wakulla) were not predicted well by any of the forecasting methodologies. Further, the forecasts for several counties (e.g., Leon, Flagler, Hardee, Okeechobee) were consistently too high or too low. These deviations indicate that there may be other influences on a county's construction activity that were not accounted for in our research. These variances may be potential opportunities for future research. Construction-Market Classiflcation hi Chapter 4, cluster analysis was described as a way to evaluate the relationships between the clusters of counties and each forecasting methodology. The R and standard of the estimate was computed between the five county clusters and the output variables from the forecasting methodologies. Due to the overall consistency in the annual trends of the forecast methodologies, and the scope limits of our research, the cluster analysis was only completed on the latest 2002 variable data set. The findings of the forecast methods and cluster comparison are provided in Table 6-6. Table 6-6 lists the forecast methodologies down the colunms, and the clusters across the rows. The counties included in each of the five clusters are shown along with the corresponding and standard error of the estimate. The overall average R^ of the forecast methodologies was found to be relatively high in clusters 2 through 5 with coefficients ranging from a high of 0.882 to a low of 0.707. The average R^ for all forecast methodologies were below average (0.527) in county cluster 1 .

PAGE 123

109 The direct forecast using residential permit activity (RPERMIT) produced the lowest correlation across all of the clusters with an average of only 0.462. The second lowest forecast method across the county clusters was construction payroll (CPAYROLL) with an average of 0.741. The R^ for all of the other forecast methods across the clusters were above 0.762.

PAGE 124

no s is p arisen. TOTEMPLY com] luster OPE T( id cl ISO O OINDEX thodo] met ting forecast its of 1 0 Resu] a a t o Tab] Cluiter O O! " « 'S °3 ^ a I ° ^ e (S M M ^ = s i m 6 ta 2 2 a & Ml 3 B-JS -i 5 5-3 ^ > p o a 3 g> V & a u eo S H I 11. 1 1 ^ & O O H -S J d w ^ 1^ 3 s " a B I I 2 (5 (rt o o o ^ <0 o o o 00 in ^ P So ^ V Ov o d o O C« O 00 O .-i O M-l <7> O o o o o o o o o o o o o tS P= ^ ^ § -8 » " s 11 illllrl ^ g :s = 2 =

PAGE 125

CHAPTER 7 CONCLUSIONS OF THE STUDY The final chapter of our research presents the conclusions of our study. This chapter begins with a review of the research intent and expectations and is followed by a summary of the research results. The overall threats to the research validity are discussed. Finally, several opportunities are identified regarding future extensions and uses of the presented research. Research Intent and Expectations The general domain of our research was market forecasting in the U.S. construction industry. More specifically, our research focused on large design and construction firms working in the nonresidenfial construction-market. Our research was intended to initiate decisions regarding the spatial structure of an organization and was used to locate new potential markets. A unique new forecasting method was presented that integrates Sir Isaac Newton's natural science theory of momentum. This forecasting approach along with several other existing approaches was used to estimate the nonresidential construction-market activity throughout the State of Florida. Issues leading to our research included unusually sudden changes in the construction industry caused by the recent economic recession combined with the market fallout fi-om the events of September 1 1, 2001. Despite the magnitude of these historic events, large D&C firms were soon pressured by their investors to return to the growth they knew before 2001. These investor expectations drove the search for new markets. Ill

PAGE 126

112 Many questions arise when searching for new markets including; where are the best new markets? What are the key indicators that best predict construction activity? What methodologies are available to identify these new markets? Are complicated forecasting methods really more accurate than a simple, more direct, approach? Which forecasting methods are the most accurate? Finally, how should the different construction-markets be segregated so the highest potential markets can be prioritized and pursued? Momentum Theory One of the most important accomplishments of our research was the introduction of the momentum forecasting theory. Our research used the logic of Newton's momentum theory to successfully derive a new forecasting methodology and apply it to strategic construction-market forecasting. The two most significant differences between this new momentum methodology and traditional statistical regression approaches include; (a) variable dimensions, and (b) multivariate measurement (see Chapter 3). Finally, the procedure for applying and analyzing county momentum was detailed. This procedure included; calculating the mass, velocity and influence of the key indicators; calculating the key indicator momentum and total county momentum; and calculating the momentum index and its corresponding slope. Comparative Validation of the Forecasting Methods In addition to the new forecasting method above, five alternative forecasting methods were used to forecast nonresidential construction activity and to comparatively validate the new momentum forecasting method. This provided a total of six forecasting methods that were used in our research. These six methods included; 1) momentum

PAGE 127

113 analysis, 2) direct forecasting, 3) factor analysis, 4) multivariate-regression, 5) MERIC economic analysis, and 6) gap analysis. These six methods produced a total of twelve individual forecasts. Momentimi analysis produced two forecasts (index and index slope), the direct method produced six forecasts, and the remaining four methodologies produced four corresponding forecasts. A one, two and three year trend projection analysis was completed for all of the forecasting methods to test for any inherent advantages relating to the length of the trend forecast. The forecasted results from the six methods were compared against a county's actual construction activity using simple regression. During this regression analysis, the MERIC and gap methodologies consistently demonstrated a weaker correlation to county's actual construction activity and were excluded from further use in the research. Next, all 67 Florida counties were rank ordered using the remaining ten forecasts and compared to their actual rank order. Finally, three classification techniques were used to cluster and compare the counties based on the characteristics of the key indicators. These three classification techniques included cluster regression analysis, gap analysis, and a county's share of the total nonresidential market. Research Results Key Construction Indicators Over 250 potential variables were identified during the literature review. A total of sixteen key construction indicators were initially selected for use in our research and were previously outlined in both Chapter 2 and 3. Eleven of the sixteen variables were highly correlated to the dependent variable nonresidential permit (NRPERMIT). These

PAGE 128

114 eleven variables had Pearson's Correlation Coefficient averages ranging from 0.829 to 0.962. The remaining five variables were subsequently dropped from use in the research due to their low correlation to NRPERMIT. Total annual tax collections by or within a county (TOTTAX), total annual assessed value of conmiercial land in a county (ALVALUE), and the total annual number of wage and salary jobs in a county (TOTEMPLY) had the first, second, and third highest correlations respectively with the dependent variable NRPERMIT. An unexpected outcome of our research was the relatively low correlation of a county's total population (POP) with NRPERMIT. The POP variable finished with only the V"' best correlation. The results of the remaining variables were outlined in Chapter 5. Next, the analysis of the three construction-market classification methodologies showed low between-method correlations ranging from 0.647 to 0.762. While these correlations may appear relatively low, it should be understood that these different classification methods are fiindamentally different measures by the nature of their variables. As was shown in Chapter 5, simultaneously viewing the counties through all three of these classification filters provides advantages over the use of just a single classification method. Finally, our research evaluated the relationships between the key construction indicators and the county clusters. The overall average correlation of the key indicators was found to be highest in clusters of counties with the largest average key indicator values. All key indicators correlated poorly in the county cluster that had the lowest average key indicator values. Therefore, our research found that the counties with the highest key indicator values are the most predictable in regards to nonresidential

PAGE 129

115 construction activity. These findings were further validated by the variance analysis in Chapter 6. Forecasting Methods The average values of for all the forecast methodologies were .796, .790, and .783 for the one, two, and three year trend projections respectively. As expected, this decrease in the average values indicates a weaker relationship as the length of the trend projections increase. The average values for the standard error of the estimate were approximately $169, $179, and $191 million for the 1, 2, and 3 year trend projection respectively. This again confirms a weaker predictive relationship as the length of the trend projections increase. With the exception of the MERIC methodology, the large F statistic values indicate that overall the methodology output variables strongly explain the variation of the independent variable. Similarly, large / values indicate the correlation between the dependent and independent variables are statistically significant, and are not caused by sampling variability. Methodology vs. complexity The methodology direct forecasting was found to be the best predictor of nonresidential construction activity in our research. A direct forecast using the TOTTAX variable consistently provided the best forecast for the 1, 2, and 3 year trend projections. Direct forecasts are the simplest of the forecasting methods used in our research. Probably the biggest surprise of our research was the direct forecast using a county's residential permit activity (RPERMIT). RPERMIT was found to be the least accurate of the six direct forecasts and provided only the 10'*" best overall forecast for the 1, 2, and 3 year trend projections. Starting in 1996, Florida county governments were no

PAGE 130

116 longer required to collect and report on total and nonresidential building permit data. It was thought that residential permit data was an accurate reflection of the overall construction activity. While RPERMIT does correlate well to total building permit activity (R^ approximately 0.925), residential building permit activity actually has a relatively weak correlation to nonresidential building permit activity (R^ approximately 0.729). These findings are concerning because nonresidential building permits accounted for over 45% of the total construction permits authorized in the State of Florida from 1990 through 2002. Our research has found that residential permit data is one of the least accurate ways of predicting nonresidential building activity in the State of Florida relative to the other methods tested. The second most accurate type of forecasting method was factor analysis. Factor analysis finished 2"*^ on average for the 1 , 2, and 3 year trend projections and was the second most complex forecasting methodology used in our research next to multivariateregression. The next most accurate type of forecasting method was multivariate-regression. Multivariate-regression provided the 3"^ most accurate forecast on average and was the most complex of all the forecasting methodologies used in our research. The next most accurate forecasting method was the new momentum analysis index and its corresponding slope. Both momentum analysis forecasts consistently finished 5* and 6"^. Momentum analysis was designed to be less complex but more meaningfiil than the traditional statistical regression approaches. Based on the accuracy of its prediction (R^ ranged fi-om 0.897 to 0.909), this new momentum methodology clearly did a good

PAGE 131

117 job of forecasting a county's construction activity. Ideas on how to improve the accuracy of this methodology were discussed in chapter 6. One of the most inaccurate forecasting methods was gap analysis. Gap analysis consistently finished with the second worst prediction for the 1, 2, and 3 year trend projections. While gap analysis did not perform well in predicting the level of construction activity in a county, this methodology was successfully used as one of the tools in our research to cluster and prioritize the 67 Florida counties. Finally, the most inaccurate forecasting method was the MERIC economic momentum analysis. MERIC 's economic momentum methodology consistently finished with the worst prediction of construction activity. It should be noted that the MERIC index variables were originally intended to be a measure of the economic momentum in a county relative to the overall economic momentum of the state. The MERIC index was not intended to measure a specific sector of the economy (i.e., construction) within a county. Best construction-markets Based on the findings of our research, a direct forecast using a county's total tax revenue (TOTTAX) was consistently found to be the best predictor of nonresidential construction activity. The top five nonresidential construction-markets in the State of Florida for the years 2004 and 2005 are forecasted to be; 1) Dade, 2) Broward, 3) Orange, 4) Palm Beach, and 5) Hillsborough counties. The top 15 counties were very consistent in their rank throughout the study period. The counties ranked through 40"" were also consistent in their rankings with some exceptions. The remainder of the counties demonstrated an overall increase in rank fluctuation.

PAGE 132

118 As expected, there is an overall increase in the total variance for the 1, 2, and 3 years trend projections. This confirms that the overall variance in the forecasts increase as the length of the trend projection increases. Our research also found that as a county's construction activity increases, the variance in a county's forecasted rank decreases. This provides further evidence that a county with a larger market share is more predictable, and a county with a smaller market share is less predictable. Threats to Research Validity The following sections of this chapter review the possible threats to the validity of our research. These threats include external, construct, statistical and internal threats.' External Validity The largest threat to the external validity of our research design appears to be interaction of the causal relationship with the units. An effect found at the county level may not hold at a lower city level or a higher state or national level. Further, the effect found with Florida counties may not hold if larger counties like those found in the Western states were used. The next threat to the external validity of our research design appears to be interaction of the causal relationship with its settings. An effect found in the tourism economy of Florida may not hold in the industrial economy of New Jersey, or the agricultural economy of Kansas. Further, an effect found in a high growth economy may not hold in a region with negative economic growth. ' Shadish, William R. et al. (2001). Experimental and Quasi-Experimental Designs for Generalized Causal Inference. Boston Massachusetts: Houghton Mifflin, (p. 38).

PAGE 133

119 Construct validity The largest threat to the construct validity of our research design appears to be an inadequate explication of variable constructs. The research design included sixteen independent variables grouped into six constructs. Not thoroughly describing or understanding all of these variable constructs may result in incorrect conclusions about the relationships between the operation and the construct. The variables used in the constructs have to be chosen specifically for the type of construction-market sector under study. An example of this is the effect found in the nonresidential construction-market segment may not be the same in a more general total construction-market segment because the total market includes residential construction. Also, an effect found in the educational construction-market segment would not be the same in the transportation or environmental construction-market segments. Further, an effect found in the general educational construction-market segment may not be the same in a more specific elementary or university construction-market segment. The next largest threat to the construct validity of our research could be construct confounding. Depending on the type or level of construction-market sector under study, other constructs such as political or envirormiental risk may need to be evaluated. Not including all of the possible variable constructs may result in incomplete construct inferences. Another threat to construct validity could be mono-operation bias. The construct variables may not thoroughly represent the variable construct or may represent more than one construct. Not including all of the possible variables may result in incomplete variable constructs.

PAGE 134

120 Statistical Validity The largest threat to the statistical validity of our research design appears to be the unreliability of measures. Due to truncated data sets, variables such as permit activity were obtained from multiple sources. Any difference or error in measurement may unknowingly weaken or strengthen the relationships between the variables. The next threat to the statistical validity of our research may be inaccurate effect size estimation. As discussed in Chapter 6, the three momentum theory variables (mass, velocity and influence) are all weighted equally. This may systematically overestimate or underestimate the size of the effect and subsequently the accuracy of the forecast. Another threat to the statistical validity of our research may include an improper cutoff point for removing variables with low correlation. Excluding certain variables may result in improper inferences. Internal Validity The largest threat to the internal validity of our research design appears to be ambiguous temporal precedence. Our research predicted the dependent variable using the independent variable values from the same year. But changes in the dependent variable may be caused by prior changes in the independent variables. An example of this relationship is as follows: An increase in population in Year-1; may spur housing starts and increase gross sales in Year-2; will increase tax revenue in Year-3; which ultimately may drives nonresidential construction activity. A lack of clarity about which variable occurred first may cause confiision about which variable is the cause and which is the effect. This is an opportunity for improvement on fixture research.

PAGE 135

121 The next threat to the internal validity could be instrumentation. The nature, or way, the variables (e.g., permits) were measured by the counties may have changed over time. This change could be confused with the variable effect. Future Research The following paragraphs in this chapter outline several future opportunities for improvement to our research. Key Indicators The best opportunity to enhance the key indicators is to experiment with the +250 different independent variables identified in Chapter 2 and Appendix A. The goal would be to improve the prediction model and subsequent forecasts. Another idea is to research a specific construction-market segment such as transportation and select the dependent and independent variables that are specific to that industry. Similarly, the research could be completed in different geographic regions with economies that are driven by the same industries as Florida, or different industries such as manufacturing. Another idea for future research could include using a national index such as the U.S. Department of Labor's Producer Price Index or Consumer Price Index, instead of using a single dimensional variable. Also, more research could be done with regional utility agencies to find better local variables to measure initial infrastructure investment. Environmental and Political Influences While the literature review combined with the attached appendices are evidence that a thorough review has been completed in regards to the selection of the variables used in our research, more experimentation can be done. Our research is by nature quantitative. As such, the variables included in our research are consistent with the economic forecasting variables traditionally used by the construction industry to forecast

PAGE 136

122 construction activity. But, it would be interesting and intriguing to add political or natural environment variables into our research. Variables measuring local environmental concerns, political and social attitudes regarding growth, or related construction industry regulation may add a unique and interesting dimension to the research outcome. Momentum Forecasting Theory The best opportunity to enhance the new momentum forecasting methodology is to experiment with the calculation of the momentum index itself The idea is to minimize the prediction error which may be accumulating error from the previous forecasts. Another idea is to investigate how to minimize the burden of low correlating variables. A higher threshold of Pearson's Correlation may need to be set so better predicting variables are included. The next opportunity relates to the variable dimensions. As previously discussed, more research needs to be done regarding the appropriate weights of each dimension of the momentum variable. Construction-Marketplace Life Cycle Products, like people, have been viewed as having a life cycle. Berkowitz et al. discusses the concept of the product life cycle as the four stages a new product goes through in the marketplace.^ These four stages include; introduction, growth, maturity and decline. Berkowitz et al. indicate that as the marketing objectives of an organization change during each stage of the product life cycle, the Marketing Mix must also change. The marketing mix is defined by Berkowitz et al. as the marketing manager's controllable factors, the marketing actions that he or she can take to solve a marketing ^ Berkowitz, E.N. et al. (2000). (p. 314).

PAGE 137

123 problem.^ These four marketing mix factors, or the Four Ps, were first published by Professor E. Jerome McCarthy (1960)"* and include; product, price, promotion and place. The findings of our research suggest that the construction-marketplace itself may go through a similar life cycle and that a large D&C firm's marketing objectives and marketing mix must be adjusted accordingly. The stages of product life cycle and the marketing mix factors outlined by Berkowitz et al. have been adapted and are presented in Table?1. Table 7-1. The relationship between the four construction-marketplace life cycle stages and marketing mix actions.^ Four Life Cycle Stages Actions Introduction Growth Mature Decline Marketing Objective Monitor Gain awareness, brand recognition Stress differentiation, maintain brand loyalty Harvesting, deletion Competition None local Growing Many Reduced Product Isolated services only Cornerstone services Full services Best selling services Price Minimal profit Penetration, gain share Defend share, profit Stay Profitable Promotion Respond to RFP's only Inform and educate Stress competitive differences Reminder oriented, minimal Place (Distribution) Temporary project offices Convert temporary office to permanent Maximize presence Consolidate offices ^ Berkowitz, E.N. et al. (2000). (p. 13). * McCarthy, E. J. (1960). Basic Marketing: A Managerial Approach. Homewood, IL: Richard D. Irwin; and Walter van Waterschoot & Christopher Van den Bulte. (October 1992). The 4P Classification of the Market Mix Revisited. Journal of Marketing, (pp. 83-93). ' Adaptation from Berkowitz, E. N. et al. (2000). Figure 12-1. (p. 315).

PAGE 138

124 A future opportunity for research could include archival research on Florida's mature construction-markets (e.g., Hillsborough [Tampa], or Dade [Miami]). The time period for this study would be from a county's introduction stage through to its current mature market life cycle stage. These counties would have theoretically been through several development stages. The variables that have historically most influenced the county's construction activity during the different stages of the lifecycle could be identified. The thirteen year research period used for this previous research was not long enough to test this marketplace lifecycle concept. A potential challenge for this research idea is that Hillsborough and Dade counties may have developed over a period of +100 years. Data availability is a key element to the success of this future research. Other Opportunities Our research could also be applied at a regional or state level and be used to support a more general theory of U.S. construction activity development. The research methodologies could also be applied to international construction-market research if an evaluation of risk was included. Further, the collection of key construction indicator data could be automated and analyzed using a geographic information system. This would allow a national D«&C firm to continually monitor the nations 3,099 counties creating a more integrated plan for long-term market expansion. It may also indicate the need to partner with, or acquire, a local design or construction firm in a high opportunity construction-market.

PAGE 139

APPENDIX A LIST OF POTENTIAL KEY CONSTRUCTION INDICATORS

PAGE 140

126 (/I o p o Q X O O 00 g lU o J2 b ^ o > U 00 U _ E a o O IG o k P. OS OS oT OS o o OS OS I o > u >s > 3 o u p." Q 1/3 D ' B-, 3 O CA , o , a! § 3 a o a. o o o — ' & c ex 2 ^ o o E Z " Q 00 05 o ^ >s >. .P J3 O O >s >s U (U to > > < < CO 3 3 s pq pq o _ 5b g O w O 3

PAGE 141

o 00 B OS Q o V 1— ^ CO o 00 a t)-i o o U 00 3 o § Q I > 1/2 .5 ^ p E •s -a Ji ° ^ s O. SO U XI S « (U M "J « "> 3 « c= g.s (DUO g § -^^ .52 . ' *^ 3 60 S > < o "a. E w I "rt 60 H > U O (u 0C u §

PAGE 142

m o 00 Q > U o 00 X u 00 a. o (3 •a § ? §1 to ^ s S i g =2 w § • . W CO c ^ ^ S 6 " ? to r^ S li) 2 "> 5 a 6 -S ^ o o a: f2 ffi 3 O X

PAGE 143

129 u u o B Q 00 u a o CO ea T3 U o ea u g e o. o o iC a (L> a 00 a, I ca 60 S — eg < I la aa ca ea > a

PAGE 144

130 U T3 O O o CO M ^ § E a •S ^ (11 -tJ C3 2P " C 60 l-E 2 3 60
PAGE 145

131 p o 1/3 B Q fl 1-1 O O 00 o •a o ca (L> o c4 g 6 ex o ea o «C G 1) u ^ O 'S > U 0) ca i3 3 .b 3 o o ca c 2 III II g ta 3 u o ^ >-J 4J «l O o IS u
PAGE 146

132 B u o "3 o 00 dex crease •pop measu u cx o u O cn 3 -4-) o ;ale crea evio U ° §-1 c .2 'E g « .2 ao. o to u y -S O J) c o 2 ^ § W 3 •£ b o u <)-( tt-i o o o. § c o o 60 ^ « u a o D a: u lis 1 (A U < z w a

PAGE 147

133 o on B Q X O a a o > U pu, w ^ a on C3 -o u o g E cx o o C > 'o u u Ia. a. w 60 c "o o o > o 3 U ^.2 " a ^ ^ cd <^ C a; U H Uh — o S a. O 3 X> B E N E S I g 3 O o o o o pL, 00 g1 Oh C

PAGE 148

134 o on B Q o on o •a D c > 'S > U CD TD
PAGE 149

135 p o B P X u o <3 O 00 E i2 o ca U

PAGE 150

136 o o i Q u o o u (5J o CO I •c > H > U •ti ^ ^ 1 1I i 2 X w > S " o > o o <5 fsj O VD O O O (N II d •5 .S -a .S-§ .>< .S 8 o. u I u u "3 O o u 'i 1 i (A C« ca ca U u u i i c/l t/5 u 3 ca O 3 3 'C O li g o c o C3 o •f D. 6 6 o e i 3 3 1 O 3 o 3 (73 3 ca ca > o c S o 5 o EC C Oh o o c o I ca M > ca u u ^ i . g c ^ Z cH (2 •s .g CO U T3 'S o g s s o o 1 6 § > 0) o g O 00 o <^-^ O u 60 , u 3 o u o ta g o § 4= 3 O e 3 s -8 o § E Iu 1 § I u o u u 60 ca < o o OS o o o o 4= 1 E E o § o 3 cu o u ta u 1 CO ca ca u g 3 § o ii .H J= ca o S *-» -4-* w (i> di " ^ s o u « "5 u u •R o +-> CO 3 O T3

PAGE 151

r 137 Hi o U I <; Oi I "J > U IC g o o to ^ T3 § b s o 2 S ^io o o P 2 O > < -S ^ •§) b ^ .2, c II II

PAGE 152

138 o o Q X o CO •a > Xi o 13 M > O T3 3 > g .. a -c to ^ so P. o o a. E Q o -3 £ . §>l " > T3 p w fa 1> c > ^ C(J ^ .S c o S W 00 G x> 2 I > o u o u p. to o u o E o •+1 o as I •o P > .sue > i ca ^ S _y >< 2 flj ™ tJ C o «i 5J w " ^ «i S II .1 2 £ 60 o o o IC u IC u s u u O. cd T5 u go IS if

PAGE 153

139 u o o 3 P X u 5 o CO a o •a o o 0 u o s c u U oj in & o cd o e II o " W M V g 13 > X o (4-. o c o g IS o O Q o — o « o c C V o — o A II te' •o u > 3 o T3 s > It o A E C3 o 0. U 1 £? o c tt-i o § I o U 3 T3 c 3 O I o o o 'S o c o g o c o U E. ?2 =5 > § o I o o o t2 T3 § o .2 o 3 O" o C '3 u a 3 CT < u § o u p. b I 3 c. g XI O o u

PAGE 154

140 u o o I p X (Zl o 3 ^ o (3 60 > U •X3 4-» 3 X) ID it: u O CA o Q D 60 .g I'S o a " o 2 3 (U o 60 a u >< > w § 1 T3 O is u o E ^ u v 60 60 § H jC P o -a Q. " •SSc CO O 8° t tiPQ " .e3 a o u CO • u o § u 3 o i |1 U w (U o o g g o u y ^ u D. ^ O d. g CO o o > H CQ
PAGE 155

141 o 00 B eit Q u C4 O C/3 o •a > 'S > U Oh 00 g J' 3 w < o u I o U E « p 'a P i anit anil u u 00 60 lar lar lar Vi 1 1 or) !/5 00 u 60 IS •a o o o CA O II 0 ^ o 09 <« ^ o es •8 *-» o IS o u u is 1) 13 .S g o IS II CO CD U ^ "J ^ V3 ^ o to g T3 O CO ci PL, ^

PAGE 156

142 o B CO Q o CO o 00 (ft a> a > ^ o CO OO > O O u CJ CO 0> C/] (U

PAGE 157

143 o % 00 +-» > u

PAGE 158

144 ' Detailed descriptions and specific applications of the variables listed in Appendix A can be obtained from the following variable sources. • City of Cincinnati. (2002, 10/01/02). Quality of Life Index. Retrieved November 17, 2003, from http://www.cincinnati-oh.gov/cmgr/downloads/cmgr_pdf4396.pdf • City of Pasadena Public Health Department. (2002). Pasadena / Altadena Quality of Life Index. Retrieved November 17, 2003, from http://www.ci.pasadena.ca.us/publichealth/qualityOfLife/QOLI 2002.pdf • Construction and Real Estate Market Pulse. Steele Analytics. Retrieved October 1, 2003, from http://www.steeleanalytics.com/construction.htm • Economic Trends and Commercial Construction Indicators for Metropolitan Washington. Metropolitan Washington Council of Governments. Retrieved October 1, 2003, from http://www.mwcog.org/uploads/committee-documents/9FtYXw200307 1 5 144 1 12.ppt • Erickson, P. A. (1994). A Practical Guide to Environmental Impact Assessment. San Diego: Academic Press, (pp. 7-29). • Fetterhoff, O. (1992). Siting Waste Facilities Using Geographic Information Systems. University of Buffalo, Buffalo, New York. (pp. 10, Appendix A). • Finkel, G. (1997). The economics of the construction industry. Armonk, N.Y.: M.E. Sharpe. (pp. 21, 24, 28-31, 34,36, 44, 57-57, 60, 64-71). • Grand Traverse Regional Community Foundation. (1998). Q1998 Quality of Life Index for the Grand Traverse Region. Retrieved November 17, 2003, from http://qualityindex.nmc.edu/ • Gujarati, D. (1984). Government and Business. New York: McGraw-Hill Book Coirrpany. (pp.2147). • Jain, R. K. et al. (2002). Environmental Assessment. New York: McGraw-Hill. (pp. 126-128, 407565, 580). • Marriott, B. B. (1997). Environmental Impact assessment, A Practical Guide. New York: McGrawHill. • New Jersey Construction Reporter, March 2003 Highlights. Division of Codes and Standards, New Jersey Department of Community Affairs. Retrieved October 1, 2003, from http://www.state.nj.us/dca/codes/cr/subform.shtml. • Rogue Valley Civic League. (2000). Southern Oregon Quality of Life Index. Retrieved November 17, 2003, from http://www.sou.edu/sorsi/RVCL ResList.PDF • Smyth, H., & NetLibrary Inc. (1999). (pp. 74, 140-141). • Social Planning Council. (2002, 03/26/02). The Quality of Life in Kingston and Area. Retrieved November 17, 2003, from http://www.spckingston.ca/Quality%20ofyo20Life%20Index.htm • Society for Marketing Professional Services. (2000). Marketing handbook for the design & construction professional. Los Angeles: BNI Building News. (pp. 37-38). • Standard & Poor's DRI & F.W. Dodge. ( 1999, December), (pp. 1 1-13).

PAGE 159

145 Terpstra, V., & David, K. (1985). The Cultural Environment of International Business. Cincinnati: SouthWestern Publishing Company. (p.61, 132, 147-152, 192-196, 199-204). Thompson, A. R. (1985). Economics. Reading, MA: AddisonWesley Publishing Company, (pp. 116-120). Thomsen, C. (1989). (p .35). Wright, R. (2002, October 16). U.S. Construction Activity stagnant, but Promises Gradual Improvement; Government-related Construction Shows Diminished Activity in 2003. North American Construction Forecast. Retrieved October 1, 2003, from http://www.nacf.com/simonson 02.html.

PAGE 160

APPENDIX B RESEARCH VARIABLE DATA FOR YEARS 1990 THROUGH 2002

PAGE 161

CO •a > 23 147 ^IlipiiippPiiiillliiPiiiiiiiiiiiii^lliii^ilP^piiP IP|i5iii|lp§lillPiililllp|Psi|||i||||pii|^ iiilissiliiilllsllilisliigsissiisiiiliipi Illlllllilll^llllllllllliillliiiilillllllliw iiPiiiiiiliiiiiiSliiiiiiilpiliiilPiiiilliiiitiliiiiliiiiiilii p^||^||i|l|iiiliiPSiSi||l|i|P5i||Pi||p^ iijiijij^iiiiiiiiiiiiiiiiiiliki^.iiiijiiiiijij^^

PAGE 162

»— H On OS CN 148 §liippippPiiillillipiiiiiiiiiiiii^liPi^i|p^pil^ §Piii||lPiB§|iPII^BI§l|5|§||l§i||i?l|Pllli|l^ liliiiPPI|i|P|||5P5Biilil|l|pl|||P^^^ pisiiiiilliliisiilliiliillilsllsllilliliiliis^ PPIi^i^llli5^|i?sP?lll|i|piPI||PBI|||||Pp||^ liill^lliiillli^illlilliliiiiliillil^^^^^ iPiilllpi|lit|iiiiiiii|iiiiiiiliiililli|i|iiiiiiilliiiPii ii^ii^lllispiiiP^§SBiiiliiip5^i||ii^ll|pPii|lililP|ii^ ^PiPiiiP|PiiPPi3ii§ii|iii§gillP^lllii|i|ipi li S'l||llli'l|j!|| iiiiiiiiiiiiiliijii3.jrfiiijjji!ijyijiiiij«jiij

PAGE 163

ON I m 149 ^Ilipi^ippPiiiilPIPPiliiiiiiiilii^llPi^iiP^P^ iPiB||li||ppE||ililBIIB|||piill|||l55|||||||||||i ^|lp||lillilpi|ilil&sa|||||pilP^ pilillllilllillllllllliilillilllllllllllllllll^ |iii|plii!liiiiiiipiipi|ipiiil|lilslii||iii|l|iililililiiiiiii l§B||3|§||l|lipil|s=PII|ipiPB|PII^^^^ Pil|P||||l|gS||||l35lBS§|i|ipil|||PS|||||^ |lil|p§|||i|IB|||gilglSI||||l||l!|||Pi|||||p^ si iJllliidlsiiilillliililllllillliijilllliljjjilL^

PAGE 164

150 ||pilipi§ip|iPiiiilPlPP|li^iiPiiliiilliipilP iPP|piiiii|i§iiPP§ii£iii|5ip5i|P^^iiiiiii|i|^ lil|l|pill|l|lijillpiBiip|ipi|||pii^^^^^ il|||il||ililil||||Iii||E|||il|||li||||||||||s^ 03 1 On PQ 1sii!iii!iipi|liiPiiiiiiii|iii|liiii|iiiiP|lilliiiililiii|ii Pp||§|iiPi^^|iliP^§§§||i|l||IS||||l|||||||||^ li.lllJJsriliiiillJllidil|}|j|jll3.iilllj|jj||il|j!iiil|i|iii|jj

PAGE 165

151 ^liippippPiiillliiipiliiiliiiiliiiliiii^ilpipiipi^ipp iiii^|PiliP|P|iPP^5iBiliplii^l||P^I|lll|l||||||^ |illi|B||||i|li|pBp?iiii|||i|ii5|||Pi|||^ 5lliiis|3|i|ii|igi=|^i|||||i||iliiPiliP l|l||iii|||||g|pip!l§l||ipiPS|||llipill|S^ lllllilililllillliillliililll^lllilllllllllllllllli iiii|Piiiiiiiiiii|iii|iiiiiiiiiiiiiiiiiiiiiji!iiiiiiiiiiii|iii fi|ggiPiiipgigp|liisiiiii§ii|||sg|||pg||gpgggg^ SPi|§|i||5|IE||ilii|PPIi|gi|l^||PII|||ip||p ^l5|pill|ii^§|ilS|i§^BI!||p|il5|||pi|||||P||p m

PAGE 166

ON ffl 152 ||Pi|ipiiippiiiiiilliiipiiiliiisiiiiii|iiiisi|p^^ P|l||s||||||P|||liBB5lll|||l|P^|||PI|||||||p ^|l|l|plil|lpi|p5|iiilil|l|iilP^^^ ^Piplllll|l§|ili|B5li||iPIil5i||Bg||« iilliilllllilil^lililil^lllilliiillllilllil^ iiiiiiliiliiiilPiiiiiiiiiiiPiilliiilliiiliillliiliilliiip [iiigapiiiigfpiiiiiiiifiii8|iiiig§iiipi|pp i§||3i|gP|l5||ilPIESi|||pE|S|||p ilil||!|||l|B§|i|J|EI§Bliiip|lll|||PI|^ p§||Sili|l|PI|llii?i^iiil|§|PB|l|P^||||l|i^

PAGE 167

> On ON PQ 153 iiliippipppiiiligliipilliiiiiH PP||5ii||||P||PP5Ell||||l|i5l|H ^ii|i|jlili|lilipllilBlill|lp|IPi||i jPil||5||||||P|||i|lll^l|iipiil^ llllllllllllliliillliillllllilliiilli iii||lilil!i§l||ii!iiiili|i|liiii|||ilili!i!P|iji|iil^ i|§||pg|l||||ggp|i|gpg|ll§lg||||pg|||p|||gp P S R p R a s iiii||||iiljiliilllllidllllllij!l!.iit||jllll^^

PAGE 168

154 iiilipiiipPPPiillliipiiiiliPP l|||p|||Pps||i5isllli|||p||PH lllllllilllppiplililiillllpp iilsiiiiiii3p5il2sS|i5|Sii|2i |iP|Plilil|l5|piP§llli||}l|i5l|||lll^ illlllilililililililllllllilillillil^ 1 as 00 ^illilliilPlliliiliPiPi^ f s § s Si S -a s s S 3 8

PAGE 169

155 00 ON llPliipPippPiiiiPIPPillllli B||ip|||pjP||PilllB||iip|iBi|^ l|||i|pp||ipi||||SI|ili§|i|li||B|^ ll|i||lilliipi||PillsllillP|P llllliilllllillllilllllllilli iil|ii||ij!iiiiiiliHllil|ipiiii|ilii!i|ii|pii !pP|p|iliii-liiii5§i§^iiippB§llii^i||is^ Ib IIIIIs ''Lso^lsills Istlllallll 6'i|sgg5i|»l"<,!' §^-^^gl

PAGE 170

On On ON 156 ||iii|ipPippi^iii|||iiipillili i|l|i||lilip|is||Pl5B|lli||p|B ^ii|ili^iii|lpi||iiiiiiiil|lpiiiiili ^iiiliisilili^slisllisilisssiiliiiliisi teil||l|||||pi|iii|5i35l||||l||il|l|3^ llililllililllillillllliiiilliilil^ ||p55||§sl53l|S = |ll^ = « = SS|5|§p 1ii||iiiiiipii!ilii||||lil|lipliiiil||ii iiiillJliiiiiiiilillliijifIlijyi^«irf|jjiij|^^

PAGE 171

157 ||piii|iiiip|iPPiil||iiP|liiii ip|l|p||||ipi|i|slslll|||ipiP^^ llilllplllllpiplllillllllillllpip iiii|p||||l|§i||is|ii3iiiiipiP5i|p^ llilllilililllilllllillllililllli lil|piiiii|iiii|lii|lill!iili|il|iliiii|iiiiliiii i|||fss|8!|t|!iSS|iiiiii|giSi!|ii|sssrsi!esiii|ssstg8g|e|H > o o o ipi = iPiiiiS|g = |ii5551332sppp5 = ^ !|iP||BiliPpBi|P|IEillligppil||^ IjiillJIisiljiiilillliiliillllliHs^jifllijjillilijlilijIllilujij

PAGE 172

1— H O o 158 lp|||p|||||pi|||lilBII||iiP|PB^ llllllpllllillllpililllllilp iiiiipisippiiiiiiiiiiiiiiii^ ipplpillpplp llllllilllllilllililllllllllili '|ii||liii|ipipplili!ilii|iiiilii!il|iiii|ii||lil|li !S||p£!8ii|t8ti!||S!!S!PSI|S(|!||!Eg||p|||||pSS«8ISg|^ !|Pi:5|i;|i;56|55j| = = = :|3 = 5S5 = |S555^ l|ii5||5Pip|iS||i|Piilll|||J|iSi|||i5i|||||^ I|tilllljllj|lllllllll3llllllillll3sjl!llllill!llijlsjlilllllf|^^

PAGE 173

159 1 O o m illipiiipPPPiiyiPPiiiiil^ ipp|P|ii|ipEi|PPEIEI|iip ^illliPlllllliBpllliililiip^^ |l|l|piiilipi|||BP5lllli||^ lllililillliillllllllilllili iSpiiS|||p|SSp.5SISSS§igg|P|g§55||SSS^ liliPliliiilllllipilllilllliliiPiliiiiiiip illfllggplgllggplllgllllgppgllggllglllglgl^ iii||iliiljiliill!lllij|fllllilll^«i!llllli|j^

PAGE 174

LIST OF REFERENCES Adler, L. (1967). Systems Approach to Marketing, Harvard Business Review. Sept./Oct. Answers to Frequently Asked Questions. Trendcast, LLC. Retrieved October 13, 2003, from http://www.trendcast.com/ama/ing/faq.htm. Armstrong, J. S. (1978). Long-Range Forecasting: From Crystal Ball to Computer. New Yorfc Wiley. Associated General Contractors of America., & AGC Construction Marketing Committee. (1995). A Contractor's Guide to Focus Sales and Increase Profitability for the Associated General Contractors of America: A Marketing Workbook for Contractors. Washington, D.C.: Associated General Contractors of America. Bennett, P. D. Dictionary of Marketing Terms (2"*^ ed.). Lincolnwood, IL: NTC Publishing Group, 1995. Berkowitz, E. N., Kerin, R. N., Hartley, S. W., and Rudelius, W. (2000). Marketing (6th ed.). Boston Massachusetts: Irwin McGraw-Hill. BNI Building News Society for Marketing Professional Services. (2000). Marketing Handbook for the Design & Construction Professional. Los Angeles: BNI Building News. BrainyMedia.com. Wayne Gretzky Quotes. Retrieved February 17, 2004, from http://www.brainyquote.eom/quotes/authors/w/wayne_gretzky.html. Brallier, J. (2002). Who Was Albert Einstein? (1st ed.). New York: Grosset & Dunlap. Brown, B. H. (1974). An Econometric Forecasting Model for a Segment of the Construction Market. Unpublished Ph.D. Dissertation, Oklahoma State University, Stillwater. Bureau of Economic and Business Research. (2002). Florida Statistical Abstract 2002. Gainesville: University of Florida, Warrington College of Business Administration. Capeci, J. D. & Campillo, M. (April 2002). Global Sector Momentum in the Emerging Markets. Cambridge, MA: Arrow Street Capital. City of Cincinnati. (August 10, 2002). Quality of Life Index. Retrieved November 17, 2003, from http://www.cincinnati-oh.gov/cmgr/downloads/cmgr_pdf4396.pdf. 160

PAGE 175

161 City of Pasadena Public Health Department. (2002). Pasadena / Altadena Quality of Life Index. Retrieved November 17, 2003, from http://www.ci.pasadena.ca.us/publichealth/qualityOfLife/QOLI_2002.pdf. Clapp, J. M. (1987). Handbook for Real Estate Market Analysis. Englewood Chffs, NJ: Prentice-Hall, Inc. Clapp, J. M. (1993). Dynamics of Office Markets. Washington, DC: The Urban Institute Press. Clark, H. (2002). Smart Momentum, the Future of Predictive Analysis in the Financial Markets. Chichester: John Wiley Sc Sons, LTD. Construction and Real Estate Market Pulse. Steele Analytics. Retrieved October 1, 2003, from http -.//www. steeleanalytics.com/construction.htm. Development Education Program Web, Glossary. The World Bank Group. Retrieved October 08, 2003, from http://www.worldbank.org/depweb/english/modules/glossary.html. Economic Indicators. Missouri Economic Research and Information Center. Retrieved October 13,2003, from http://www.ded.state.mo.us/business/researchandplanning/indicators/momentum/in dex.shtinl. Economic Trends and Commercial Construction Indicators for Metropolitan Washington. Metropolitan Washington Council of Governments. Retrieved October 1, 2003, from http://www.mwcog.org/uploads/committeedocuments/9FtYXw2003071 5 1 441 12.ppt. Erickson, P. A. (1994). A Practical Guide to Environmental Impact Assessment San Diego: Academic Press. Fetterhoff, O. (1992). Siting Waste Facilities Using Geographic Information Systems. Unpublished Master of Science Thesis, University of Buffalo, Buffalo, New York. Finkel, G. (1997). The Economics of the Construction Industry^ Armonk, N.Y.: M.E. Sharpe. Fisher, N. (1986). Marketing for the Construction Industry: A Practical Handbook for Consultants, Contractors, and other Professionals. London: Longman ; J. Wiley. Friedman, W. (1984). Construction Marketing and Strategic Planning. New York: McGraw-HilL Gerwick, B. C, & Woolery, J. C. (1983). Construction and Engineering Marketing for Major Project Services. New York: Wiley.

PAGE 176

161 Giancoli, D. C. (1980). Physics. Englewood Cliffs, NJ: Prentice-Hall, Tnc Grand Traverse Regional Community Foundation. (1998). Q1998 Quality of Life Index for the Grand Traverse Region. Retrieved November 17, 2003, from http://qualityindex.nmc.edu/. Granger, C. W. J., & Newbold, P. (1977). Forecasting Economic Time Series. New York: Academic Press. Gujarati, D. (1984). Government and Business. New York: McGraw-Hill Book Company. Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate Data Analysis (5th ed.J. Upper Saddle River, New Jersey: Prentice Hall. Hillebrandt, P. M. (1974). Economic Theory and the Construction Industry. London: Macraillan. Hillebrandt, P. M., & Cannon, J. (1990). The Modern Construction Firm. London: Macmillan. Huff, D. L. (1964, July). Defining and Estimating a Trade Area. Journal of Marketing. Jain, R. K., Urban, L. V., Staccy, G. S., Balbach, H. E., & Webb, M. D. (2002). Environmental Assessment. New York: McGraw-Hill. Levenbach, H., & Cleary, J. P. (1984). The Modern Forecaster: The Forecasting Process through Data Analysis. Belmont, Calif.: Lifetime Learning Publications. Marriott, B. B. (1997). Environmental Impact Assessment, A Practical Guide. New York: McGraw-Hill. McCarthy, E. J. (1960). Basic Marketing: A Managerial Approach. Homewood, IL: Richard D. Irwin; and Walter van Waterschoot & Christopher Van den Bulte. (October 1992). The 4P Classification of the Market Mix Revisited. Journal of Marketing. McGraw-Hill Companies, C. L G. (1999, December). World Market Overview. Engineering News Record. MerriamWebster Online Dictionary. Retrieved February 2, 2004, from http://www.mw.coffl/cgl-bin/dietionary Momentum Analysis. Momentum Analysis Opinion Research, LLC. Retrieved October 08, 2003, from http://www.ballot.org/resources/Momentum_Analysis_Brochure.doc.

PAGE 177

163 Momentum Indicators. Keystone Commodity Trading Guide. Retrieved October 13, 2003, from http://www.keystone-web.com/technicals/momentum.html. Montgomery, D. C, Johnson, L. A., & Gardiner, J. S. (1990). Forecasting and Time Series Analysis (2nd@d.). New York: McGraw-Hill. National Environmental PoUcy Act, Title I, Sec 102(2) (A). Hew Jersey Construction Reporter, March 2003 HighUghts. Division of Codes and Standards, New Jersey Department of Community Affairs. Retrieved October 1, 2003, from http://www,state.nj.us/dca/codes/cr/subform.shtml. Newton, Isaac. (1687). Principia, Translated by Andrew Motte 1729. Organisation for Economic Co-operation and Development. (2002). Main Economic Indicators: Comparative Methodological Analysis : Industry, Retail and Construction Indicators. Paris: Organisation for Economic Co-operation and Development. Pickar, R. L., AGC Construction Marketing Committee. (1995). A Contractor's Guide to Focus Sales and Increase Profitability: A Marketing Workbook for Contractors. Washington, D.C.: Associated General Contractors of America. Pring, M. (1993). Martin Pringon Market Momentum. Sarasota, FL: International Institute for Economic Research, Inc. Pring, M. (2002). Momentum Explained. New York: McGraw-Hill. Reilly, J. W. (1959). Methods for the Study of Retail Relationships. Austin, TX: University of Texas. Rogue Valley Civic League. (2000). Southern Oregon Quality of Life Index. Retrieved November 17^ 2003, from http://www.sou.edu/sorsi/RVCL_ResList.PDF. Shadish, W. R., Thomas D. Cook, and Donald T. Campbell. (2001). Experimental and Quasi-Experimental Designs for Generalized Causal Inference. Boston Massachusetts: Houghton Mifflin. Smyth, H. J. (1998). Innovative Ways of Segmenting the Market: Practice Guide No. 1. Oxford: Oxford Brookes University, Center for Construction Marketing. Smyth, H., & NetLibrary hic. (1999). Marketing and Selling Construction Services. Maiden, Mass.: Blackwell Science. Social Planning Council. (2002, 03/26/02). The Quality of Life in Kingston and Area. Retrieved November 17, 2003, from http://www.spckingston.ca/Quality%20oP/o20Life%20hidex.htm.

PAGE 178

IM Society for Marketing Professional Services. (2000). Marketing handbook for the design & construction professional. Los Angeles: BNI Building News. Standard & Poor's, DRl & F.W. Dodge. (1999, December). Building New markets: Global Construction Market Opportunities and Risks (Prospectus for Multi-Client Study). Lexington, MA: The McGraw-Hill Companies Construction Information Group. Statistical Package for the Social Sciences, Inc. Base 10.0 Applications Guide (1999). Chicago, IL:SPSS Inc. Stochastic Momentum Index. Paritech Inc. Retrieved October 13, 2003, from http://www.paritech.com/education/technical/indicators/momentum/stochasticl.asp Stock Market Momentum Indicator. Commodity Research Bureau. Retrieved October 13, 2003, from http://www.crbtrader.com/crbindex/nsmmi.asp. Terpstra, V., & David, K. (1985). The Cultural Environment of International Business. Cincinnati: SouthWestern Publishing Company. Thompson, A. R. (1985). Economics. Reading, MA: AddisonWesley Publishing Company, Thomsen, C. (1989). Managing Brainpower: Organizing, Measuring Performance, and Selling in Architecture, Engineering, and Construction Management Companies. Washington, D.C.: American Institute of Architects Press. TradingSolutions Function Library, Relative Momentum Index [RMI]. Trading Solutions. Retrieved October 13, 2003, from http://www.tradingsolutions.com/functions/RelativeMomentumIndex.html. Tulacz, G., & Powers, M. (May 19, 2003). The Top 400 Contractors. ENR. Value/Momentum Sector Analysis August 3 1 , 2001. Ford Equity Research. Retrieved October 13, 2003, from http://www.fordequity.com/html/documents/Ss0801.pdf Wright, R. (2002, October 16). U.S. Construction Activity stagnant, but Promises Gradual Improvement; Government-related Construction Shows Diminished Activity in 2003. North American Construction Forecast. Retrieved October 1, 2003, from http://www.nacf com/simonson_02. html.

PAGE 179

BIOGRAPHICAL SKETCH Otto G. Fetterhoff III was bom on November 12*, 1964, in Niagara Falls, New York, and was raised in the small upstate town of Sanborn, New York. Otto's construction education and experience began as a child in a family-owned generalcontracting construction firm. For over 30 years, Fetterhoff Construction Corporation provided commercial construction services in upstate New York. Otto graduated from Erie Community College in Buffalo, New York, with an Associate in Applied Science degree in construction technology. Otto then attended Rochester Institute of Technology (RIT) and graduated with a bachelor's degree in civil engineering. While at RIT, Otto worked as a professor's assistant in the soils and foundations laboratory, and also in the environmental controls laboratory. Otto went on to complete his graduate education at the State University of New York, University of Buffalo. In 1992, Otto graduated with a Master of Science degree in the social sciences focusing on economics and public administration. His final graduate project applied the newly emerging geographic information systems technologies to siting waste disposal facilities. In 1991, Otto accepted a position with the Rust International and Waste Management family of companies. Otto also taught part-time as an adjunct professor at RIT, in the College of Engineering, in 1995 and 1996. In 1996, Otto accepted a position with the Walt Disney World Company in Orlando, Florida, and relocated to Central 165

PAGE 180

166 Florida. In 2000, Otto returned to the Consulting side of the construction industry, and accepted a position with the URS Corporation, his current employer. Otto returned to academia as a part-time student to earn his Doctor of Philosophy degree from the University of Florida in 2000. Otto completed his requirements for candidacy in 2003. Otto currently lives in Montverde, Florida, with his wife Michele, and their two sons Hans and Alexander.

PAGE 181

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. William J. O^Brien, Chair Assistant Professor of Building Construction I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. Marc T. Smith, Cochair Associate Professor of Building Construction I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully ad^ate, in scope and quality, as a^ dissertation for the degree of Doctor of Philosophy. Charles J. Kibert I Professor of Building Construction I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy. David C. Ling Professor of Finance, Insurance, and Real Estate I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality, as a dissertation for the degree of Doctor of Philosophy Robert C. Stroh, Sr. Lecturer of Building Construction

PAGE 182

This dissertation was submitted to the Graduate FacuUy of the College of Design, Construction and Planning and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy. May 2004 Dean, College(blf Planning sign. Construction and Dean, Graduate School